Microsoft Research Colloquium

Microsoft Research Colloquium


The Microsoft Research Colloquium at Microsoft Research New England focuses on research in the foundational aspects of computer science, mathematics, economics, anthropology and sociology. With an interdisciplinary flavor, this colloquium series features some of the foremost researchers in their fields talking about their research, breakthroughs and advances.

The agenda typically consists of approximately 50 minutes of prepared presentation and brief Q&A, followed immediately by a brief reception* to meet the speaker and address detailed questions. We welcome members of the local academic community to attend.



Upcoming Speakers

Capitalism and Entrepreneurship in Socialist China

Adam Frost, Harvard
Wednesday, May 08, 2019
4:00 PM – 5:00 PM


Contrary to popular belief, the modern Chinese economy did not spring into being in 1978 with Deng Xiaoping’s call for “Reform and Opening Up.” Rather, it was a product of an incremental, bottom-up transformation, decades in the making. As my research shows, throughout China’s socialist era, citizens at all levels of society— from farmers who illegally traded ration coupons, to state officials who colluded with underground factories to manufacture goods— actively subverted state control to profit from inefficiencies in planning and, more generally, to make things work. In the absence of “good institutions,” they formed illicit networks that subsumed the ordinary functions of markets (ex. coordination, information aggregation, risk sharing, lending) and created productive assemblages of capital, labor, and knowledge. Drawing upon an array of unconventional sources that have never before been examined by scholars, I will argue that capitalism and entrepreneurship not only supported the functioning of China’s socialist economy, but fundamentally reshaped it and created the conditions for subsequent economic growth.


Adam Frost is a Ph.D. student at Harvard University who specializes in the economic history of modern China. His research broadly explores the history of informal economies, from unlicensed taxi drivers in 1920’s Shanghai to underground entrepreneurs in socialist China to beggars in contemporary Xi’an. In addition to conducting traditional archival research, Adam draws heavily upon ethnography, oral history, and contraband documents. Most recently he completed a documentary on the everyday lives of beggars in Northwest China entitled, The End of Bitterness.

Nancy Baym & Hunt Allcott, MSR- NE
Wednesday, June 19, 2019
4:00 PM – 5:00 PM

Shiri Azenkot, Cornell
Wednesday, July 10, 2019
4:00 PM – 5:00 PM

Past Speakers

New Algorithms for Interpretable Machine Learning in High Stakes Decisions- Cynthia Rudin

Cynthia Rudin, Duke| Wednesday, December 12, 2018


With widespread use of machine learning, there have been serious societal consequences from using black box models for high-stakes decisions, including flawed models for medical imaging, and poor bail and parole decisions in criminal justice. Explanations for black box models are not reliable, and can be misleading. If we use interpretable models, they come with their own explanations, which are faithful to what the model actually computes. I will present work on (i) optimal decision lists, (ii) interpretable neural networks for computer vision, and (iii) optimal scoring systems (sparse linear models with integer coefficients). In our applications, we have always been able to achieve interpretable models with the same accuracy as black box models.


Cynthia Rudin is an associate professor of computer science, electrical and computer engineering, and statistics at Duke University, and directs the Prediction Analysis Lab, whose main focus is in interpretable machine learning. Previously, Prof. Rudin held positions at MIT, Columbia, and NYU. She holds an undergraduate degree from the University at Buffalo, and a PhD in applied and computational mathematics from Princeton University. She is the recipient of the 2013 and 2016 INFORMS Innovative Applications in Analytics Awards, an NSF CAREER award, was named as one of the “Top 40 Under 40” by Poets and Quants in 2015, and was named by as one of the 12 most impressive professors at MIT in 2015. Work from her lab has won 10 best paper awards in the last 5 years. She is past chair of the INFORMS Data Mining Section, and is currently chair of the Statistical Learning and Data Science section of the American Statistical Association. She has served on committees for DARPA, the National Institute of Justice, and AAAI. She has served on three committees for the National Academy of Sciences, including the Committee on Applied and Theoretical Statistics, the Committee on Law and Justice, and the Committee on Analytic Research Foundations for the Next-Generation Electric Grid.

Will AI Cure Healthcare?- Ernest Fraenkel

Ernest Fraenkel, MIT| Wednesday, October 03, 2018


Artificial intelligence (AI) is widely touted as the solution to almost every problem in society.AI is predicted to transform the workplace, manufacturing, farming, marketing, banking, insurance, transportation, policing, education and even dating. What are the prospects for applying AI to healthcare? What problems are ripe for data driven approaches? Which solutions are within reach if we plan properly, and which remain in the distant future?  I will provide somewhat opinionated answers to these questions and look forward to a healthy discussion.


Ernest Fraenkel is a Professor of Biological Engineering at the Massachusetts Institute of Technology. His laboratory seeks to understand diseases from the perspective of systems biology.T hey develop computational and experimental approaches for finding new therapeutic strategies by analyzing molecular networks, clinical and behavioral data. He received his PhD in Biology from MIT after graduating summa cum laude from Harvard College with an AB in Chemistry and Physics.

Biased data, biased predictions, and disparate impacts: Evaluating risk assessment instruments in criminal justice- Alex Chouldechova

Alex Chouldechova, Carnegie Mellon University| Wednesday, August 29, 2018


Risk assessment tools are widely used around the country to inform decision making within the criminal justice system. Recently, considerable attention has been paid to whether such tools may suffer from predictive racial bias, and whether their use may result in racially disparate impact. Evaluating a tool for predictive bias typically entails a comparison of different predictive accuracy metrics across racial groups. Problematically, such evaluations are conducted with respect to target variables that may represent biased measurements of an unobserved outcome of more central interest. For instance, while it would be desirable to predict whether an individual will commit a future crime (reoffend), we only observe proxy outcomes such as rearrest and reconviction.  My talk will focus on how this issue of “target variable bias” affects evaluations of a tool’s predictive bias.  I will also discuss various reasons why risk assessment tools may result in racially disparate impact.


Alexandra Chouldechova is an Assistant Professor of Statistics and Public Policy at Carnegie Mellon University’s Heinz College.  Her research over the past few years has centered on fairness in predictive modeling, particularly in the context of criminal justice and public services applications.  Some of her recent work has delved into tradeoffs between different notions of predictive bias, disparate impact of data-driven decision-making, and inferential challenges stemming from common forms of “data bias”.  A statistician by training, Alex received her Ph.D. in Statistics from Stanford University in 2014.

Delegating Computation- Yael Kalai

Yael Kalai, MSR-NE| Wednesday, August 22, 2018


Efficient verification of computation, also known as delegation of computation, is one of the most fundamental notions in computer science, and in particular it lies at the heart of the P vs. NP question. In this talk I will give a brief overview of the evolution of proofs in computer science, and show how this evolution is instrumental to solving the problem of delegating computation. I will highlight a curious connection between the problem of delegating computation and the notion of no-signaling strategies from quantum physics.


Most recently an Assistant Professor of Computer Science at Georgia Tech. Before this, Yael was a post-doc at the Weizmann Institute in Israel and Microsoft Research in Redmond. She graduated from MIT, working in cryptography under the superb supervision of Shafi Goldwasser.

The Dynamics of Network Formation and Some Social and Economic Consequences- Matt Jackson

Matt Jackson, Stanford| Wednesday, August 15, 2018


Technological advances are changing interaction patterns from world trade to social network patterns. Two different implications of evolving networks are discussed – one is changing trade patterns and their impact on military alliances and wars, and the other is the formation and evolution of friendships among students, and resulting academic performance.


Matthew O. Jackson is the William D. Eberle Professor of Economics at Stanford University and an external faculty member of the Santa Fe Institute and a senior fellow of CIFAR.  He was at Northwestern University and Caltech before joining Stanford, and received his BA from Princeton University and PhD from Stanford.  Jackson’s research interests include game theory, microeconomic theory, and the study of social and economic networks, on which he has published many articles and the books Social and Economic Networks and The Human Network.  He teaches an online course on networks and co-teaches two others on game theory.  Jackson is a Member of the National Academy of Sciences, a Fellow of the American Academy of Arts and Sciences,  a Fellow of the Econometric Society, a Game Theory Society Fellow, and an Economic Theory Fellow, and his other honors include the von Neumann Award of the Rajk Laszlo College, a Guggenheim Fellowship, the Social Choice and Welfare Prize, and the B.E.Press Arrow Prize.

The Welfare Effects of Information- Cass Sunstein

Cass Sunstein, Harvard| Wednesday, August 08, 2018


Some information is beneficial; it makes people’s lives go better. Some information is harmful; it makes people’s lives go worse. Some information has no welfare effects at all; people neither gain nor lose from it. Under prevailing executive orders, agencies must investigate the welfare effects of information by reference to cost-benefit analysis. Federal agencies have (1) claimed that quantification of benefits is essentially impossible; (2) engaged in “break-even analysis”; (3) projected various endpoints, such as health benefits or purely economic savings; and (4) relied on private willingness-to-pay for the relevant information. All of these approaches run into serious objections. With respect to (4), people may lack the information that would permit them to say how much they would pay for (more) information; they may not know the welfare effects of information; and their tastes and values may shift over time, in part as a result of information. These points suggest the need to take the willingness-to-pay criterion with many grains of salt, and to learn more about the actual effects of information, and of the behavioral changes produced by information, on people’s experienced well-being.


Cass R. Sunstein is currently the Robert Walmsley University Professor at Harvard. From 2009 to 2012, he was Administrator of the White House Office of Information and Regulatory Affairs. He is the founder and director of the Program on Behavioral Economics and Public Policy at Harvard Law School. Mr. Sunstein has testified before congressional committees on many subjects, and he has been involved in constitution-making and law reform activities in a number of nations.

Mr. Sunstein is author of many articles and books, including (2001), Risk and Reason (2002), Why Societies Need Dissent (2003), The Second Bill of Rights (2004), Laws of Fear: Beyond the Precautionary Principle (2005), Worst-Case Scenarios (2001), Nudge: Improving Decisions about Health, Wealth, and Happiness (with Richard H. Thaler, 2008), Simpler: The Future of Government (2013) and most recently Why Nudge? (2014) and Conspiracy Theories and Other Dangerous Ideas (2014). He is now working on group decision making and various projects on the idea of liberty.

Planar Graph Perfect Matching is in NC- Vijay Vazirani

Vijay Vazirani, University of California, Irvine | Wednesday, August 01, 2018 | Video

Please Note: This seminar is of a more technical nature than our typical colloquium talks.


Is matching in NC, i.e., is there a deterministic fast parallel algorithm for it? This has been an outstanding open question in TCS for over three decades, ever since the discovery of Random NC matching algorithms. Within this question, the case of planar graphs has remained an enigma: On the one hand, counting the number of perfect matchings is far harder than finding one (the former is #P-complete and the latter is in P), and on the other, for planar graphs, counting has long been known to be in NC whereas finding one has resisted a solution!

The case of bipartite planar graphs was solved by Miller and Naor in 1989 via a flow-based algorithm.  In 2000, Mahajan and Varadarajan gave an elegant way of using counting matchings to finding one, hence giving a different NC algorithm.  However, non-bipartite planar graphs still didn’t yield: the stumbling block being odd tight cuts.  Interestingly enough, these are also a key to the solution: a balanced tight odd cut leads to a straight-forward divide and conquer NC algorithm. However, a number of ideas are needed to find such a cut in NC; the central one being an NC algorithm for finding a face of the perfect matching polytope at which Omega(n) new conditions, involving constraints of the polytope, are simultaneously satisfied.

Paper available at:

Joint work with Nima Anari.


Vijay Vazirani received his BS at MIT and his Ph.D. from University of California, Berkeley. He is currently Distinguished Professor at University of California, Irvine.  He has made seminal contributions to the theory of algorithms, in particular to the classical maximum matching problem, approximation algorithms, and complexity theory. Over the last decade and a half, he has contributed widely to an algorithmic study of economics and game theory.

Vazirani is author of a definitive book on Approximation Algorithms, published in 2001, and translated into Japanese, Polish, French and Chinese. He was McKay Fellow at U. C. Berkeley in Spring 2002, and Distinguished SISL Visitor at Caltech during 2011-12. He is a Guggenheim Fellow and an ACM Fellow.

Can Intricate Structure Occur by Accident?- Henry Cohn

Henry Cohn, MSR-NE | Wednesday, July 25, 2018


Many topics in science and engineering involve a delicate interplay between order and disorder.  For example, this occurs in the study of interacting particle systems, as well as related problems such as designing error-correcting codes for noisy communication channels.  Some solutions of these optimization problems exhibit beautiful long-range order while others are amorphous.  Finding a clear basis for this dichotomy is a fundamental mathematical problem, sometimes called the crystallization problem.  It’s natural to assume that any occurrence of dramatic structure must happen for a good reason, but is that really true?  I wish I knew.  In this talk (intended to be accessible to an interdisciplinary audience) we’ll take a look at some test cases.


Henry Cohn’s mathematical interests include symmetry and exceptional structures; more generally, he enjoys any area in which concrete problems are connected in surprising ways with abstract mathematics.  He came to MSR as a postdoc in 2000 and joined the theory group long-term in 2001.  In 2007 he became head of the cryptography group, and in 2008 he moved to Cambridge with Jennifer Chayes and Christian Borgs to help set up Microsoft Research New England. He stays up late at night worrying about why the 16th dimension isn’t like the 8th or 24th.

The Simple Economics of Artificial Intelligence- Avi Goldfarb

Avi Goldfarb, University of Toronto | Wednesday, July 18, 2018


Recent excitement in artificial intelligence has been driven by advances in machine learning. In this sense, AI is a prediction technology. It uses data you have to fill in information you don’t have. These advances can be seen as a drop in the cost of prediction. This framing generates powerful, but easy-to-understand implications. As the cost of something falls, we will do more of it. Cheap prediction means more prediction. Also, as the cost of something falls, it affects the value of other things. As machine prediction gets cheap, human prediction becomes less valuable while data and human judgment become more valuable. Business models that are constrained by uncertainty can be transformed, and organizations with an abundance of data and a good sense of judgment have an advantage.

Based on the book Prediction Machines by Ajay Agrawal, Joshua Gans, and Avi Goldfarb.


Avi Goldfarb is the Rotman Professor of Artificial Intelligence and Healthcare at the Rotman School of Management, University of Toronto, and coauthor of the Globe & Mail bestselling book Prediction Machines: The Simple Economics of Artificial Intelligence. Avi is also Senior Editor at Marketing Science, Chief Data Scientist at the Creative Destruction Lab, and Research Associate at the National Bureau of Economic Research where he helps run the initiatives around digitization and artificial intelligence. Avi’s research focuses on the opportunities and challenges of the digital economy. He has published over 60 academic articles in a variety of outlets in marketing, statistics, law, computing, management, and economics. This work has been discussed in White House reports, Congressional testimony, European Commission documents, The Economist, Globe and Mail, National Public Radio, The Atlantic, New York Times, Financial Times, Wall Street Journal, and elsewhere. He holds a Ph.D. in economics of Northwestern University.

You Can Lead a Horse to Water: Spatial Learning and Path Dependence in Consumer Search- Greg Lewis

Greg Lewis, MSR- NE | Wednesday, July 11, 2018 | Video


We introduce a model of search by imperfectly informed consumers with unit demand. Consumers learn spatially: sampling the payoff to one product causes them to update their payoffs about all products that are nearby in some attribute space.  Search is costly, and so consumers face a trade-off between “exploring” far apart regions of the attribute space and “exploiting” the areas they already know they like. We present evidence of spatial learning in data on online camera purchases, as consumers who sample unexpectedly low quality products tend to subsequently sample products that are far away in attribute space. We develop a flexible parametric specification of the model where consumer utility is sampled as a Gaussian process and use it to estimate demand in the camera data using Markov Chain Monte Carlo (MCMC) methods. We conclude with a counterfactual experiment in which we manipulate the initial product shown to a consumer, finding that a bad initial experience can lead to early termination of search. Product search rankings can therefore substantially affect consumer search paths and purchase decisions.


Greg Lewis is an economist, whose main research interests lie in industrial organization, market design and applied econometrics. He received his bachelor’s degree in economics and statistics from the University of the Witwatersrand in South Africa, and his MA and PhD both from the University of Michigan.  He then served on the economics faculty at Harvard, as assistant and then associate professor.  Recently, his time has been spent analyzing strategic learning by firms in the British electricity market, suggesting randomized mechanisms for price discrimination in online display advertising, developing econometric models of auction markets, and evaluating the design of procurement auctions.

New Media, New Work, and the New Call to Intimacy: The Case of Musicians- Nancy Baym

Nancy Baym MSR- NE | Wednesday, June 27, 2018


The architectures and norms of new media push people toward sharing everyday intimacies they might historically have kept to close friends and family. As more people are pushed toward gig work, the original gig workers – musicians – provide an exemplary lens for exploring the implications of this widespread blurring of interpersonal communication into everyday practices of professional viability. This talk, based on the new book Playing to the Crowd: Musicians, Audiences, and the Intimate Work of Connection, draws on nearly a decade of work to show how the pressure to be “authentic” in communicating with audiences, combined with the designs and materialities of new communication technologies, raises dialectic tensions that musicians – and many others – must manage as social media platforms become integral to professional life.


Nancy Baym is a Principal Researcher at Microsoft Research in Cambridge, Massachusetts. After earning her Ph.D. in 1994 in the Department of Speech Communication at the University of Illinois, she was on the faculty of Communication departments for 18 years before joining Microsoft in 2012. With Steve Jones (and others), she was a founder of the Association of Internet Researchers and served as its second President. She is the author of Personal Connections in the Digital Age (Polity Press), now in its second edition, Tune In, Log On: Soaps, Fandom and Online Community (Sage Press), and co-editor of Internet Inquiry: Conversations About Method (Sage Press) with Annette Markham. She serves on the editorial boards of New Media & Society, the Journal of Communication, The Journal of Computer Mediated Communication, and numerous other journals. Her book Playing to the Crowd: Musicians, Audiences, and the Intimate Work of Connection will be published in July by NYU Press.  More information, most of her articles, and some of her talks are available at

Custodians of the Internet: Platforms, Content Moderation, and the Hidden Decisions that Shape Social Media- Tarleton Gillespie

Tarleton Gillespie, MSR- NE | Wednesday, June 13, 2018


This talk will give an overview of my new book, and highlight the public debate about content moderation and its implications for those studying or building information systems that host user content.

Most social media users want their chosen platforms free from harassment and porn. But they also want to see the content they choose to see. This means platforms face an irreconcilable contradiction: while platforms promise an open space for participation and community, every one of them imposes rules of some kind. In the early days of social media, content moderation was hidden away, even disavowed. But the illusion of the open platform has, in recent years, begun to crumble. Today, content moderation has never been more important, or more controversial. In this book, I discuss how social media platforms police what we post online – and the societal impact of these decisions. Content moderation still receives too little public scrutiny. How and why platforms moderate can shape societal norms and alter the contours of public discourse, cultural production, and the fabric of society — and the very fat of moderation should change how we understand what platforms are.


Tarleton Gillespie is a principal researcher at Microsoft Research, an affiliated associate professor in Cornell’s Department of Communication and Department of Information Science, co-founder of the blog Culture Digitally, author of Wired Shut: Copyright and the Shape of Digital Culture (MIT, 2007), co-editor of Media Technologies: Essays on Communication, Materiality, and Society (MIT, 2014), and the author of the forthcoming Custodians of the Internet (Yale, 2018).

Building Machines That Learn and Think Like People- Josh Tenenbaum

Josh Tenenbaum, MIT | Wednesday, June 6, 2018


Recent progress in artificial intelligence (AI) has renewed interest in building systems that learn and think like people. Many advances have come from using deep neural networks trained end-to-end in tasks such as object recognition, video games, and board games, achieving performance that equals or even beats humans in some respects. Despite their biological inspiration and performance achievements, these systems differ from human intelligence in crucial ways. We review progress in cognitive science suggesting that truly human-like learning and thinking machines will have to reach beyond current engineering trends in both what they learn, and how they learn it. Specifically, we argue that these machines should (a) build causal models of the world that support explanation and understanding, rather than merely solving pattern recognition problems; (b) ground learning in intuitive theories of physics and psychology, to support and enrich the knowledge that is learned; and (c) harness compositionality and learning-to-learn to rapidly acquire and generalize knowledge to new tasks and situations. We suggest concrete challenges and promising routes towards these goals that can combine the strengths of recent neural network advances with more structured cognitive models.


Joshua Tenenbaum is Professor of Cognitive Science and Computation at the MIT. He and his colleagues in the Computational Cognitive Science group want to understand that most elusive aspect of human intelligence: our ability to learn so much about the world, so rapidly and flexibly. While their core interests are in human learning and reasoning, they also work actively in machine learning and artificial intelligence. These two programs are inseparable: bringing machine-learning algorithms closer to the capacities of human learning should lead to more powerful AI systems as well as more powerful theoretical paradigms for understanding human cognition. Their current research explores the computational basis of many aspects of human cognition: learning concepts, judging similarity, inferring causal connections, forming perceptual representations, learning word meanings and syntactic principles in natural language, noticing coincidences and predicting the future, inferring the mental states of other people, and constructing intuitive theories of core domains, such as intuitive physics, psychology, biology, or social structure. He is known for contributions to mathematical psychology and Bayesian cognitive science. Tenenbaum previously taught at Stanford University, where he was the Wasow Visiting Fellow from October 2010-January 2011.

The Cloud is a Factory: A Material History of the Digital Economy- Nathan Ensmenger

Nathan Ensmenger, Indiana | Wednesday, May 2, 2018


Drawing on the literature on environmental history, this paper surveys the multiple ways in which humans, environment, and computing technology have been in interaction over the past several centuries. From Charles Babbage’s Difference Engine (a product of an increasingly global British maritime empire) to Herman Hollerith’s tabulating machine (designed to solve the problem of “seeing like a state” in the newly trans-continental American Republic) to the emergence of the ecological sciences and the modern petrochemical industry, information technologies have always been closely associated with the human desire to understand and manipulate their physical environment. More recently, humankind has started to realize the environmental impacts of information technology, including not only the toxic byproducts associated with their production, but also the polluting effects of the massive amounts of energy and water required by data centers at Google and Facebook (whose physicality is conveniently and deliberately camouflaged behind the disembodied, ethereal “cloud”).

More specifically, this paper will explore the global life-cycle of a digital commodity — in this case a unit of the virtual currency Bitcoin — from lithium mines in post-colonial South America to the factory city-compounds of southern China to a “server farm” in the Pacific Northwest to the “computer graveyards” outside of Agbogbloshie, Ghana. The goal is to ground the history of information technology in the material world by focusing on the relationship between “computing power” and more traditional processes of resource extraction, exchange, management, and consumption.


Nathan Ensmenger is an Associate Professor in the School of Informatics, Computing and Engineering at Indiana University, where he also serves as the Chair of the Informatics department. He specializes in the social and labor history of computing, gender and computing, and the relationship between computing and the environment.

Alan Turing: Pioneer of the Information Age- Jack Copeland

Jack Copeland, University of Canterbury | Wednesday, April 11, 2018


At the turn of the millennium Time magazine listed Alan Turing among the twentieth century’s 100 greatest minds, alongside the Wright brothers, Albert Einstein, Crick and Watson, and Alexander Fleming. Turing’s achievements during his short life of 42 years were legion. Best known as the genius who broke some of Germany’s most secret codes during the war of 1939-45, Turing was also the father of the modern computer. Today, all who click or touch to open are familiar with the impact of his ideas. To Turing we owe the concept of storing applications, and the other programs necessary for computers to do our bidding, inside the computer’s memory, ready to be opened when we wish. We take for granted that we use the same slab of hardware to shop, manage our finances, type our memoirs, play our favourite music and videos, and send instant messages across the street or around the world. Like many great ideas this one now seems as obvious as the cart and the arch, but with this single invention—the stored-program universal computer—Turing changed the world.

Turing was a theoretician’s theoretician, yet he also had immensely practical interests. In 1945 he designed a large stored-program electronic computer called the Automatic Computing Engine, or ACE. Turing’s sophisticated ACE design achieved commercial success as the English Electric Company’s DEUCE, one of the earliest electronic computers to go on the market. In those days—the first eye-blink of the Information Age—the new machines sold at a rate of no more than a dozen or so a year. But in a handful of decades, Turing’s ideas transported us from an era where ‘computer’ was the term for a human clerk who did the sums in the back office of an insurance company or science lab, into a world where many have never known life without the Internet.


Jack Copeland FRS NZ is Distinguished Professor of Philosophy at the University of Canterbury in New Zealand. He is also Co-Director and Permanent Visiting Fellow of the Turing Centre at the Swiss Federal Institute of Technology in Zurich, and Honorary Research Professor of Philosophy at the University of Queensland, Australia, and is currently the John Findlay Visiting Professor of Philosophy at Boston University.

In 2016 Jack received the international Covey Award in recognition of “a substantial record of innovative research in the field of computing and philosophy”, and in 2017 his name was added to the IT History Society Honor Roll, which the Society describes as “a listing of a select few that have made an out-of-the-ordinary contribution to the information industry”. He has just been awarded the annual Barwise Prize by the American Philosophical Association for “significant and sustained contributions to areas relevant to philosophy and computing”.

A Londoner by birth, Jack gained a D.Phil. in mathematical logic from the University of Oxford. His books include The Essential Turing (Oxford University Press); Colossus: The Secrets of Bletchley Park’s Codebreaking Computers (Oxford University Press); Alan Turing’s Electronic Brain (Oxford University Press); Computability: Turing, Gödel, Church, and Beyond (MIT Press); Logic and Reality (Oxford University Press), and Artificial Intelligence (Blackwell). He has published more than 100 journal articles on the history and philosophy of both computing and mathematical logic. In 2014 Oxford University Press published his highly accessible paperback biography Turing, and in 2017 released his latest book The Turing Guide.

Jack has been script advisor, co-writer, and scientific consultant for a number of historical documentaries. One of them, Arte TV’s The Man Who Cracked the Nazi Codes is based on his bio Turing and won the audience’s Best Documentary prize at the 2015 FIGRA European film festival; another, the BBC’s Code-Breakers: Bletchley Park’s Lost Heroes won two BAFTAs and was listed as one of the year’s three best historical documentaries at the 2013 Media Impact Awards in New York City.

Jack was Visiting Professor of Information Science at Copenhagen University in 2014-15 and Visiting Professor in Computer Science and Philosophy at the Swiss Federal Institute of Technology in 2013-14; and in 2012 he was the Royden B. Davis Visiting Chair of Interdisciplinary Studies in the Department of Psychology at Georgetown University, Washington DC.

Connected self-ownership and implications for online networks and privacy rights- Ann Cudd

Ann Cudd, BU | Wednesday, April 4, 2018


This talk explores the concept of the connected self-owner, which takes account of the metaphysical significance of relations among persons for persons’ capacities to be owners. This concept of the self-owner conflicts with the traditional, libertarian understanding of the self as atomistic or essentially separable from all others. I argue that the atomistic self cannot be a self-owner. A self-owner is a moral person with intentions, desires, thoughts. But to have intentions, desires, thoughts a being has to relate to others through language and norm-guided behavior. Individual beings require the pre-existence of norms and norm-givers to bootstrap their selves, and norms and norm-givers and norm-takers are necessary to continue to support the self. That means, I argue, that the self who can be an owner is essentially connected. Next, I ask how humans become connected selves and whether that matters morally. I distinguish among those connections that support development of valuable capacities. One such capacity is the autonomous individual. I argue that the social connections that allow the development of autonomous individuals have moral value and should be fostered; oppressive social connections, on the other hand, tend to thwart autonomy. This has implications for how we should think about privacy rights and online networks.


Ann E. Cudd is Dean of the College and Graduate School of Arts & Sciences and Professor of Philosophy at Boston University. In her role as dean, she has worked to increase diversity and inclusion at the College, fundraising tirelessly for need-based scholarships and taking concrete steps to increase faculty diversity. She has identified five areas of research and teaching excellence within the College as priorities for investment: the digital revolution in the arts & sciences, neuroscience, climate change and sustainability, the humanities, and the study of inequality. A champion of the arts and sciences, she has been instrumental in launching a dialogue among Boston colleges and universities about the importance of liberal education.

Prior to joining BU in 2015, Cudd was Vice Provost and Dean of Undergraduate Studies and University Distinguished Professor of Philosophy at the University of Kansas. Among other roles during her 27 years at KU, she directed the Women, Gender, and Sexuality Studies program and served as associate dean for the humanities in the College of Liberal Arts & Sciences. Her award-winning 2006 book, Analyzing Oppression (Oxford University Press), examines the economic, social, and psychological causes and effects of oppression. She is co-editor or co-author of six other books, including Capitalism For and Against: A feminist debate (with Nancy Holmstrom; Cambridge University Press). Her recent work concerns the moral value of capitalism, conceptions of domestic violence in international law, the injustice of educational inequality, and the self-ownership debate in liberal political philosophy. She is past president and founding member of the Society for Analytical Feminism and vice president and president-elect of the American section of the International Society for the Philosophy of Law and Social Philosophy (AMINTAPHIL). She received her BA in Mathematics and Philosophy from Swarthmore College and an MA in Economics and PhD in Philosophy from the University of Pittsburgh.

Prelaunch Demand Estimation- Juanjuan Zhang

Juanjuan Zhang, MIT | Wednesday, December 13, 2017


Demand estimation is important for new-product strategies, but is challenging in the absence of actual sales data. We develop a cost-effective method to estimate the demand of new products based on incentive-aligned choice experiments. Our premise is that there exists a structural relationship between manifested demand and the probability of consumer choice being realized. We illustrate the mechanism using a theory model, in which consumers learn their product valuation through costly effort and their effort incentive depends on the realization probability. We run a large-scale choice experiment on a mobile game platform, where we randomize the price and realization probability when selling a new product. We find reduced-form support of the theoretical prediction and the decision effort mechanism. We then estimate a structural model of consumer choice. The structural estimates allow us to infer actual demand using data from incentive-aligned choice experiments with small to moderate realization probabilities.


Juanjuan Zhang is the Epoch Foundation Professor of International Management and Professor of Marketing at the MIT Sloan School of Management.

Zhang studies marketing strategies in today’s social context. Her research covers industries such as consumer goods, social media, and healthcare, and functional areas such as product development, pricing, and sales. She has received the Frank Bass Award for the best marketing thesis, is a four-time finalist for the John Little Award for the best marketing paper, and a twice finalist for the INFORMS Society for Marketing Science Long Term Impact Award.

Zhang currently serves as Department Editor of Management Science, and Associate Editor of the Journal of Marketing Research, Marketing Science, and Quantitative Marketing and Economics. She also serves as a VP of the INFORMS Society for Marketing Science (ISMS). Zhang teaches Marketing Management at MIT Sloan. She is a recipient of the MIT d’Arbeloff Fund for Excellence in Education and MIT Sloan’s highest teaching award, the Jamieson Prize. She holds a B. Econ. from Tsinghua University and a Ph.D. in Business Administration from the University of California, Berkeley.

Are we over-testing? Using machine learning to understand physician decision making- Ziad Obermeyer

Ziad Obermeyer, Harvard | Wednesday, December 06, 2017


Low-value health care—care that provides little health benefit relative to its cost—is a central concern for policy makers. Identifying exactly which care is likely to be of low-value ex ante, however, has proven challenging.  We use a novel machine learning approach to gauge the extent of low-value care.

We focus on the decision to perform high-cost tests on emergency department (ED) patients whose symptoms suggest they might be having a heart attack—notoriously difficult to differentiate from more benign causes.  We build an algorithm to predict whether a particular patient is in fact having a heart attack using a training sample of randomly-selected patients), and apply it to a new population of patients the algorithm has never seen. We find that a large set of patients tested by doctors have extremely low ex ante predicted risk of having a heart attack; and these patients do indeed have a very low rate of positive test results when tested. Our focus on testing decisions on the margin reveals that the rate of over-testing is substantially higher than we would think if we simply measured overall effectiveness of the test: the marginal test has much lower value than the average test, and our approach can quantify this difference. We also find that many patients who go untested in fact appear high risk to the algorithm. Doctors’ decisions not to test these patients does not appear to reflect private information: we find that these patients develop serious complications (or death) at remarkably high rates in the months after emergency visits. By isolating specific conditions under which patients in emergency departments are quasi-randomly assigned to doctors, we are able to minimize the influence of unobservables. These results suggest that both under-testing and over-testing are prevalent.

We conclude with exploratory analysis of the behavioral mechanisms underlying under-testing, by examining those high-risk beneficiaries whom physicians fail to test. These patients often have concurrent health issues with symptoms similar to heart attack that may lead physicians to anchor prematurely on an alternative diagnosis. Applying deep learning to electrocardiographic waveform data from these patients, we can also isolate specific physiological characteristics of the heart attacks that doctors overlook.


Ziad Obermeyer is an Assistant Professor at Harvard Medical School and an emergency physician at the Brigham and Women’s Hospital, both in Boston.

His lab applies machine learning to solve clinical problems. As patients age and medical technology advances, the complexity of health data strains the capabilities of the human mind. Using a combination of machine learning and traditional methods, his work seeks to find hidden signal in health data, help doctors make better decisions, and drive innovations in clinical research.

He is a recipient of multiple research awards from NIH (including the Office of the Director and the National Institute on Aging) and private foundations, and a faculty affiliate at ideas42, Ariadne Labs, and the Harvard Institute for Quantitative Social Science. He holds an A.B. (magna cum laude) from Harvard and an M.Phil. from Cambridge, and worked as a consultant at McKinsey & Co. in Geneva, New Jersey, and Tokyo, before returning to Harvard for his M.D (magna cum laude).

Spatial Pricing in Ride-Sharing Networks- Kostas Bimpikis

Kostas Bimpikis, Stanford | Wednesday, November 29, 2017


We explore spatial price discrimination in the context of a ride-sharing platform that serves a network of locations. Riders are heterogeneous in terms of their destination preferences and their willingness to pay for receiving service. Drivers decide whether, when, and where to provide service so as to maximize their expected earnings, given the platform’s prices. Our findings highlight the impact of the demand pattern on the platform’s prices, profits, and the induced consumer surplus. In particular, we establish that profits and consumer surplus are maximized when the demand pattern is “balanced” across the network’s locations. In addition, we show that they both increase monotonically with the balancedness of the demand pattern (as formalized by its structural properties). Furthermore, if the demand pattern is not balanced, the platform can benefit substantially from pricing rides differently depending on the location they originate from. Finally, we consider a number of alternative pricing and compensation schemes that are commonly used in practice and explore their performance for the platform.

(joint work with Ozan Candogan and Daniela Saban)


Kostas Bimpikis is an Associate Professor of Operations, Information and Technology at Stanford University’s Graduate School of Business. Prior to joining Stanford, he spent a year as a postdoctoral research fellow at the Microsoft Research New England Lab. Professor Bimpikis has received a PhD in Operations Research from the Massachusetts Institute of Technology in 2010, an MS in Computer Science from the University of California, San Diego and a BS degree in Electrical and Computer Engineering from the National Technical University of Athens, Greece.

Matching Pennies on the Campaign Trail: An Empirical Study of Senate Elections and Media Coverage- Pinar Yildirim

Pinar Yildirim, UPENN | Wednesday, November 15, 2017


We study the strategic interaction between the media and Senate candidates during elections. While the media is instrumental for candidates to communicate with voters, candidates and media outlets have conflicting preferences over the contents of the reporting. In competitive electoral environments such as most US Senate races, this can lead to a strategic environment resembling a matching pennies game. Based on this observation, we develop a model of bipartisan races where media outlets report about candidates, and candidates make decisions on the type of constituencies to target with their statements along the campaign trail. We develop a methodology to classify news content as suggestive of the target audience of candidate speech, and show how data on media reports and poll results, together with the behavioral implications of the model, can be used to estimate its parameters. We implement this methodology on US Senatorial races for the period 1980-2012, and find that Democratic candidates have stronger incentives to target their messages towards turning out their core supporters than Republicans. We also find that the cost in swing-voter support from targeting core supporters is larger for Democrats than for Republicans. These effects balance each other, making media outlets willing to cover candidates from both parties at similar rates.

Joint work with Camilo Garcia Jimeno Department of Economics, University of Pennsylvania and NBER.


Pinar Yildirim is Assistant Professor of Marketing at the Wharton School of the University of Pennsylvania and is also a Senior Fellow at the Leonard Davis Institute.

Pinar’s research areas are media, information economics, and network science. She focuses on applied theory and applied economics problems relevant to online platforms, advertising, networks, media and political economy. Her research appeared in top management and marketing journals including Marketing Science, Journal of Marketing Research, Management Science, and Journal of Marketing. Pinar is on the editorial board of Marketing Science and recently received the 2017 MSI Young Scholar award.

She holds Ph.D. degrees in Marketing and Business Economics, as well as Industrial Engineering from the University of Pittsburgh. She joined the Wharton School in 2012 and has been teaching in the Executive, MBA, and undergraduate programs since.

Social Order in the Age of Big Data: Exploring the Knowledge Problem and the Freedom Problem– Nick Couldry

Nick Couldry, LSE | Wednesday, November 1, 2017 | Video


This talk will explore how to use social theory to understand problems of social order and its relationship to an era of Big Data. I take as a starting-point of the talk the neglected late work of theorist Norbert Elias and the concept of figurations, which I draw upon in my recent book (The Mediated Construction of Reality, with Andrew Hepp, Polity 2016), as a way of thinking better about the social world’s real complexity. I will sketch the historical background that shapes what we know about the role of communications in the growth of industrial capitalism in the 19th century to argue that we are in a parallel phase of major transformation today. This raises two problems on which the talk will reflect: first, what are the distinctive features of the social knowledge that is today being generated through big data processes, compared with the 19th century’s rise of statistics as the primary generator of social knowledge; second, what are the implications for the enduring value of freedom of the data collection processes on which Big Data is founded?


Nick Couldry is Full Professor of Media, Communications and Social Theory at the London School of Economics and Political Science, UK. From August 2014 to August 17, he was chair of LSE’s Department of Media and Communications. He is the author of 8 books, including most recently The Mediated Construction of Reality (with Andreas Hepp, Polity 2016). He has since 2015 been joint coordinating author of the chapter on media and communications in the International Panel on Social Progress: For more information, please go to

Three principles of data science: predictability, stability, and computability– Bin Yu

Bin Yu, UC Berkeley | Thursday, October 26, 2017 | Video


In this talk, I will discuss intertwining importance and connections of three principles of data science. The three principles will be demonstrated in the context of two neuroscience projects and through analytical connections. In particular, the first project adds stability to predictive models used for reconstruction of movies from fMRI brain signals to gain interpretability of the predictive models. The second project employs predictive transfer learning and stable (manifold) deep dream images to characterize the difficult V4 neurons in primate vision cortex. Our results lend support, to a certain extent, to the resemblance to a primate brain of Convolutional Neural Networks (CNNs).


Bin Yu is Chancellor’s Professor in the Departments of Statistics and of Electrical Engineering & Computer Sciences at the University of California at Berkeley. Her current research interests focus on statistics and machine learning algorithms and theory for solving high-dimensional data problems. Her lab is engaged in interdisciplinary research with scientists from genomics, neuroscience, precision medicine and political science.

She obtained her B.S. degree in Mathematics from Peking University in 1984, her M.A. and Ph.D. degrees in Statistics from the University of California at Berkeley in 1987 and 1990, respectively. She held faculty positions at the University of Wisconsin-Madison and Yale University and was a Member of Technical Staff at Bell Labs, Lucent. She was Chair of Department of Statistics at UC Berkeley from 2009 to 2012, and is a founding co-director of the Microsoft Lab on Statistics and Information Technology at Peking University, China, and Chair of the Scientific Advisory Committee of the Statistical Science Center at Peking University.

She is Member of the U.S. National Academy of Sciences and Fellow of the American Academy of Arts and Sciences. She was a Guggenheim Fellow in 2006, an Invited Speaker at ICIAM in 2011, and the Tukey Memorial Lecturer of the Bernoulli Society in 2012. She was President of IMS (Institute of Mathematical Statistics) in 2013-2014 and the Rietz Lecturer of IMS in 2016. She is a Fellow of IMS, ASA, AAAS and IEEE.

She served on the Board of Mathematics Sciences and Applications (BMSA) of NAS and as co-chair of SAMSI advisory committee, and on the Board of Trustees at ICERM and Scientific Advisory Board of IPAM. She has served or is serving on many editorial boards, including Journal of Machine Learning Research (JMLR), Annual Reviews in Statistics, Annals of Statistics, and American Statistical Association (JASA).

Consumer Reviews and Regulation: Evidence from NYC Restaurants– Chiara Farronato

Chiara Farronato, Harvard | Wednesday, October 25, 2017


We investigate complementarities and substitutabilities between two signals of restaurant quality: health inspections and online reviews. To protect consumers from unsafe dining, health inspections periodically evaluate restaurants on hygiene quality, and assign them health grades. Recently, consumers have increasingly been able to rate restaurant quality online, through platforms like Yelp. We first investigate whether online consumer reviews detect hygienic conditions that health inspectors evaluate. To do this, we implement a text analysis machine learning algorithm to predict individual restaurant violations from the text of Yelp reviews. We preliminarily find that consumer reviews are good predictors of food handling violations, but are poor predictors of facilities and maintenance violations. We then investigate how the hygienic information contained in online reviews affects consumer demand and supply incentives. On the demand side, we preliminarily find that conditional on hygiene quality contained in online reviews, customers still use health grades to choose restaurants. On the supply side, we find that relative to restaurants not on Yelp, restaurants reviewed on Yelp score better on hygiene dimensions detectable by customers than on dimensions not detectable by customers. The paper results have implications for the design of government regulation in a world where consumers rate their service experiences online.


Chiara Farronato is an assistant professor of business administration in the Technology and Operations Management Unit at Harvard Business School. Based on a broad interest in the economics of innovation and the Internet, she concentrates her research on the evolution of e-commerce and peer-to-peer online platforms, including platform adoption, economies of scale, and drivers of heterogeneous platform success. Chiara has investigated such phenomena as the shift in e-commerce from auctions to posted prices; matching supply and demand on peer-to-peer platforms for local and time-sensitive services; and the effect of peer-to-peer entry on the market structure of existing industries.

Valuing Alternative Work Arrangements– Amanda Pallais

Amanda Pallais, Harvard | Wednesday, October 11, 2017


We employ a discrete choice experiment in the employment process for a national call center to estimate the willingness to pay distribution for alternative work arrangements relative to traditional office positions. Most workers are not willing to pay for scheduling flexibility, though a tail of workers with high valuations allows for sizable compensating differentials. The average worker is willing to give up 20% of wages to avoid a schedule set by an employer on short notice, and 8% for the option to work from home. We also document that many jobseekers are inattentive, and we account for this in estimation.


Amanda Pallais is the Paul Sack Associate Professor of Political Economy and Social Studies at Harvard University. Her research studies the labor market performance and educational investment decisions of disadvantaged and socially excluded groups such as women, ethnic minorities, and individuals from low-income families. She is also interested in online platforms and how technology will change the nature of work and education.

New Frontiers in Imitation Learning– Yisong Yue

Yisong Yue, Caltech | Wednesday, September 6, 2017 | Video


The ongoing explosion of spatiotemporal tracking data has now made it possible to analyze and model fine-grained behaviors in a wide range of domains. For instance, tracking data is now being collected for every NBA basketball game with players, referees, and the ball tracked at 25 Hz, along with annotated game events such as passes, shots, and fouls. Other settings include laboratory animals, people in public spaces, professionals in settings such as operating rooms, actors speaking and performing, digital avatars in virtual environments, and even the behavior of other computational systems.In this talk, I will describe ongoing research in using imitation learning to develop predictive models of fine-grained behavior. Imitation learning is branch of machine learning that deals with learning to imitate dynamic demonstrated behavior. I will provide a high level overview of the basic problem setting, as well as specific projects in modeling laboratory animals, professional sports, speech animation, and expensive computational oracles.


Yisong Yue is an assistant professor in the Computing and Mathematical Sciences Department at the California Institute of Technology. He was previously a research scientist at Disney Research. Before that, he was a postdoctoral researcher in the Machine Learning Department and the iLab at Carnegie Mellon University. He received a Ph.D. from Cornell University and a B.S. from the University of Illinois at Urbana-Champaign.

Yisong’s research interests lie primarily in the theory and application of statistical machine learning. His research is largely centered around developing integrated learning-based approaches that can characterize complex structured and adaptive decision-making settings. Current focus areas include developing novel methods for spatiotemporal reasoning, structured prediction, interactive learning systems, and learning with humans in the loop. In the past, his research has been applied to information retrieval, recommender systems, text classification, learning from rich user interfaces, analyzing implicit human feedback, data-driven animation, behavior analysis, sports analytics, policy learning in robotics, and adaptive routing & allocation problems.

Fast Quantification of Uncertainty and Robustness with Variational Bayes – Tamara Broderick

Tamara Broderick, MIT | Wednesday, August 23, 2017 | Video


In Bayesian analysis, the posterior follows from the data and a choice of a prior and a likelihood. These choices may be somewhat subjective and reasonably vary over some range. Thus, we wish to measure the sensitivity of posterior estimates to variation in these choices. While the field of robust Bayes has been formed to address this problem, its tools are not commonly used in practice. We demonstrate that variational Bayes (VB) techniques are readily amenable to fast robustness analysis. Since VB casts posterior inference as an optimization problem, its methodology is built on the ability to calculate derivatives of posterior quantities with respect to model parameters. We use this insight to develop local prior robustness measures for mean-field variational Bayes (MFVB), a particularly popular form of VB due to its fast runtime on large data sets. A potential problem with MFVB is that it has a well-known major failing: it can severely underestimate uncertainty and provides no information about covariance. We generalize linear response methods from statistical physics to deliver accurate uncertainty estimates for MFVB—both for individual variables and coherently across variables. We call our method linear response variational Bayes (LRVB).


Tamara Broderick is the ITT Career Development Assistant Professor in the Department of Electrical Engineering and Computer Science at MIT. She is a member of the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL), the MIT Statistics and Data Science Center, and the Institute for Data, Systems, and Society (IDSS). She completed her Ph.D. in Statistics with Professor Michael I. Jordan at the University of California, Berkeley in 2014. Previously, she received an AB in Mathematics from Princeton University (2007), a Master of Advanced Study for completion of Part III of the Mathematical Tripos from the University of Cambridge (2008), an MPhil by research in Physics from the University of Cambridge (2009), and an MS in Computer Science from the University of California, Berkeley (2013). Her recent research has focused on developing and analyzing models for scalable Bayesian machine learning—especially Bayesian nonparametrics. She has been awarded a Google Faculty Research Award, the ISBA Lifetime Members Junior Researcher Award, the Savage Award (for an outstanding doctoral dissertation in Bayesian theory and methods), the Evelyn Fix Memorial Medal and Citation (for the Ph.D. student on the Berkeley campus showing the greatest promise in statistical research), the Berkeley Fellowship, an NSF Graduate Research Fellowship, a Marshall Scholarship, and the Phi Beta Kappa Prize (for the graduating Princeton senior with the highest academic average).

Music, Deep Learning, and Acoustic Invariances– Sham Kakade

Sham Kakade, University of Washington | Wednesday, August 16, 2017


Given a recorded (polyphonic) performance of, say, classical music, how can we learn to identify which notes are being played at any given time? by which instruments? if the notes are whole notes, half notes, etc.? Can we even make progress in modeling the composition process? Taking a machine learning viewpoint, one method is to learn a classifier for these tasks. Recently, we have created the MusicNet dataset to aid in the process of supervised learning on music. MusicNet consists of freely-licensed classical music recordings along with instrument/note labels, including over 40 hours of polyphonic music, covering 10 instruments and 10 composers, resulting in over 1 million temporal labels with an average of about 50 distinct notes per instrument.Given such a large scale supervised dataset, what supervised learning approaches capture the natural invariances in music? Inspired by the impressive successes of convolutional neural networks, one approach would be to train a convolutional neural network directly on the acoustic signal. We are argue (both empirically and based on the nature of acoustic signals) that this approach is lacking.  Instead, we consider a different architecture designed to capture invariances that are more natural to the way in which people perceive pitch in music (an approach which is possibly appropriate to speech recognition as well). We train this architecture in an end to end manner. Our current results already significantly outperform commercially available software for tasks such as the aforementioned one, along with the task of music transcription.

Joint work with: John Thickstun, Zaid Harchaoui, and Dean Foster.


Sham Kakade joins the University of Washington this fall with a joint position in Computer Science & Engineering and Statistics. He was most recently a principal research scientist at Microsoft Research. Prior to Microsoft, Sham held faculty positions at the University of Pennsylvania and the Toyota Technological Institute in Chicago. Sham researches artificial intelligence, statistical machine learning and signal processing, developing methods of teaching computers to make predictions based on data collection and computation time. Sham is interested in applying scalable and efficient algorithms to solve core scientific and economic problems involving complex data mining. He explores artificial neural networks that process sight and sound like the human brain and studies how machine learning intersects with neuroscience and computational biology.

The FlowNet Project: Transnational Investigations of Social Media Use and ‘Internet Freedom– Lisa Parks

Lisa Parks, MIT | Wednesday, August 9, 2017


This talk will provide an overview of findings from a three-year, interdisciplinary research project called FlowNet funded by the US State Department (2014-2017). Our team used qualitative, field-based methods to investigate social media use and internet freedom climates in Mongolia, Turkey, and Zambia. Working with local partners and translators, we interviewed nearly 200 people on the frontlines of free speech struggles in these countries — including journalists, lawyers, elected officials, LGBTQ, environmental, and social activists, media company owners, artists, and political dissidents — in an effort to understand how social media and internet freedom are culturally understood and practiced across diverse national contexts. These findings were shared with computer scientists on our team who developed an app called SecurePost, which enables anonymous, verified, group communication using existing social media platforms. The talk will conclude with a discussion of design challenges in developing contexts and the ethical considerations of developing tools and applications to support internet freedoms.


Lisa Parks is Professor of Comparative Media Studies and Director of the Global Media Technologies and Cultures Lab at MIT. Her research is focused on uses of information and media technologies across diverse, transnational cultural contexts. She is the author of Cultures in Orbit: Satellites and the Televisual (Duke UP, 2005) and Coverage: Vertical Mediation and the War on Terror (forthcoming). She is also co-editor of Life in the Age of Drone Warfare (Duke UP, 2017), Signal Traffic: Critical Studies of Media Infrastructures (U of Illinois, 2015), Down to Earth: Satellite Technologies, Industries, and Cultures (Rutgers UP, 2012), and Planet TV: A Global Television Reader (NYU Press, 2003). Parks has held visiting appointments at the Institute for Advanced Study in Berlin, University of Southern California, and the Annenberg School of Communication at the University of Pennsylvania. She has been a PI on major research grants from the National Science Foundation and the US State Department. Before joining the MIT faculty in 2017, she was Professor of Film and Media Studies at UC Santa Barbara, where she also served as the Director of the Center for Information Technology and Society (2012-2015).

Submodularity, Optimal Transport and Machine Learning: New Applications and Algorithms– Stefanie Jegelka

Stefanie Jegelka, MIT | Wednesday, August 2, 2017


Submodularity and Optimal Transport (OT) are two powerful mathematical concepts with multiple potential benefits to be exploited in machine learning. In this talk, I will outline connections and some new applications and algorithms for these concepts. First, we show how submodularity can be leveraged to solve a non-convex saddle point problem to global optimality (under conditions). The underlying method uses connections to OT. The resulting algorithm solves a robust budget allocation (or bipartite influence maximization) problem with uncertain parameters.Second, we combine submodularity and OT for a new, structured assignment model that encourages coupled assignments and has applications from domain adaptation to NLP. If time permits, I will also sketch new results on scalable computation of Wasserstein barycenters with applications to parallel Bayesian inference.

This talk is based on joint works with Matthew Staib, David Alvarez Melis, Sebastian Claici, Tommi Jaakkola and Justin Solomon.


Stefanie Jegelka is an X-Consortium Career Development Assistant Professor in the Department of EECS at MIT. She is a member of the Computer Science and AI Lab (CSAIL), the Center for Statistics and an affiliate of IDSS and ORC. Before joining MIT, she was a postdoctoral researcher at UC Berkeley, and obtained her PhD from ETH Zurich and the Max Planck Institute for Intelligent Systems. Stefanie has received an NSF CAREER Award, a DARPA Young Faculty Award, a Google research award, the German Pattern Recognition Award and a Best Paper Award at the International Conference for Machine Learning (ICML). Her research interests span the theory and practice of algorithmic machine learning.

Subgraphs, clusters and hierarchy in complex networks – Johan van Leeuwaarden

Johan van Leeuwaarden, Eindhoven University of Technology | Wednesday, July 26, 2017


Real-world networks often have power-law degrees and scale-free properties such as ultra-small distances and ultra-fast information spreading. We provide evidence of a third universal property: correlations that suppress among others the creation of triangles and signal the presence of hierarchy. We first quantify this property in terms of c(k), the probability that two neighbors of a degree-k node are neighbors themselves. We investigate how c(k) scales with k and discover a universal curve that consists of three k-ranges where c(k) remains flat, starts declining, and eventually settles on a power law with an exponent that depends on the power law of the degree distribution. We test these results against ten contemporary real-world networks (please approach us if you have your own network data set that needs to be tested) and then generalize our theory to any finite subgraph. Understanding the natural scale of all subgraphs might prove useful for community detection algorithms and establishing graph limits. Joint work with Clara Stegehuis, Remco van der Hofstad and Guido Janssen.


Johan van Leeuwaarden (1978) is professor of mathematics at Eindhoven University of Technology. He chairs the group Stochastic Networks and Applied Probability and investigates phenomena arising in complex networks, such as communication networks, social networks and biological networks, primarily through stochastic models (random graphs, interacting particles and queueing networks), in particular their scaling limits and asymptotic behavior. Johan is member of the Young Academy (part of The Royal Netherlands Academy of Arts and Sciences) and promotes the role of mathematics and data in the networked society. He also co-founded the large multidisciplinary research program NETWORKS (

Gender Salience and Racial Frames, Potholes for Women in Science: Understanding the Context Before and the Potential Consequences of Sexual Harassment– Anna Branch

Anna Branch, UMass Amherst | Wednesday, July 19, 2017 | Video


This talk grounds the experience of harassment sociologically drawing attention to the way that race and gender interact to shape who experiences gender harassment and how they respond to it. Gender salience explains how gender moves from the background to the foreground when women experience harassment and racial frames account for the hypersexualization of women of color that makes them more vulnerable to harassment. Dealing with the culture of harassment is integral to creating climates where women can fully contribute to science. Conceiving of the challenges to diversifying science as a pipeline problem, fails to appreciate the hazards, such as harassment, that women experience along the way. I introduce the road with exits, pathways, and potholes to articulates the ideas of agency and constraint for women in science and offer suggestions of what we can do to ease their journey.


Enobong Hannah Branch is an Associate Professor of Sociology and the Chancellor’s Faculty Advisor for Diversity & Inclusive Excellence at the University of Massachusetts-Amherst. Her research interests are in race, racism, and inequality; intersectional theory; work and occupations; and diversity in science. Her book Opportunity Denied: Limiting Black Women to Devalued Work (2011) provides an overview of the historical evolution of Black women’s work and the social-economic structures that have located them in particular and devalued places in the U.S. labor market. She is also the editor of Pathways, Potholes, and the Persistence of Women in Science: Reconsidering the Pipeline (2016) which outlines the inadequacy of the pipeline metaphor in understanding the challenges of entry and persistence in science and offers an alternative model that better articulates the ideas of agency, constraint, and variability along the path to scientific careers for women. Dr. Branch is also the author of several articles. Her current research, investigates rising employment insecurity in the post-industrial era through the lens of racial and gender inequality and implementing the Computer Science for All educational goals and aims within the context of existing inequality in urban school districts in Western Massachusetts.

Using Technology for Social Good: Successes, Failures, and Opportunities– Bill Thies

Bill Thies, MSR India | Thursday, July 13, 2017


While technology has offered benefits for many of us, rarely do those benefits extend equally across all members of society.  A growing community of researchers is specifically targeting their work to underserved populations, developing methods and tools that result in direct and measurable social good.  This talk will focus on those living in extreme poverty, where the latest technologies are often out of reach.  Instead of making things faster, bigger, and more futuristic, can we make things radically cheaper, simpler, and more inclusive?  I will describe some of our successes, failures, and lessons learned in deploying such “frugal technologies” in India over the past eight years.  Drawing on projects in health and citizen reporting, I will synthesize our recommendations for having social impact with technology, and outline opportunities for future work.


Bill Thies is a Senior Researcher at Microsoft Research India, where he has worked since 2008. His research focuses on building appropriate information and communication technologies that contribute to the socio-economic development of low-income communities, a field known as ICTD.  Previously, Bill earned his B.S., M.Eng., and Ph.D. degrees from MIT, where he worked on programming languages and compilers for multicore processors as well as microfluidic chips. His distinctions include the John C. Reynolds Doctoral Dissertation Award, a CHI Best Paper Award, and a 2016 MacArthur Fellowship.

Inside Job or Deep Impact? Using Extramural Citations to Assess Economic Scholarship– Josh Angrist

Josh Angrist, MIT | Wednesday, July 5, 2017


Does academic economic research produce material of broader scientific value, or are academic economists writing only for their peers? Is economics scholarship especially insular? We address these questions by quantifying interactions between economics and other disciplines. Changes in the impact of economic scholarship are measured here by the way other disciplines cite us and by the extent to which we cite others. We document a clear rise in the extramural influence of economic research, while also showing that economics is increasingly likely to reference other social sciences. A breakdown of extramural citations by economics fields shows broad field impact. Differentiating between theoretical and empirical papers classified using machine learning, we see that much of the rise in extramural influence reflects growth in citations to empirical work. This parallels a growing share of empirical cites within economics.

With Pierre Azoulay, Glenn Ellison, Ryan Hill, and Susan Lu.


Joshua Angrist is the Ford Professor of Economics at MIT, a director of MIT’s School Effectiveness and Inequality Initiative,  and a Research Associate at the National Bureau of Economic Research. A dual U.S. and Israeli citizen, he taught at Harvard and the Hebrew University of Jerusalem before coming to MIT in 1996. Angrist received his B.A. from Oberlin College in 1982 and completed his Ph.D. in Economics at Princeton in 1989.

Angrist’s research interests include the economics of education and school reform; social programs and the labor market; the effects of immigration, labor market regulation and institutions; and econometric methods for program and policy evaluation.  Angrist is a Fellow of the American Academy of Arts and Sciences and the Econometric Society, and has served on many editorial boards and as a Co-editor of the Journal of Labor Economics. He received an honorary doctorate from the University of St Gallen (Switzerland) in 2007 and is the author (with Steve Pischke) of Mostly Harmless Economics: An Empiricist’s Companion and Mastering ‘Metrics: The Path from Cause to Effect, both published by Princeton University Press. Angrist and Pischke hope to bring undergraduate econometrics instruction out of the Stones Age.

Network Pricing: How to Induce Optimal Flows Under Strategic Link Operators– Evdokia Nikolova

Evdokia Nikolova, UT Austin | Wednesday, June 21, 2017 | Video


Network pricing games provide a framework for modeling real-world settings with two types of strategic agents: users of the network and owners (operators) of the network. Owners of the network post a price for usage of the link they own; users of the network select routes based on price and level of use by other users. One challenge in these games is that there are two levels of competition: one, among the owners to abstract users to their link so as to maximize profit; and second, among users of the network to select routes that are cheap yet not too congested. Interestingly, we observe that: (i) an equilibrium may not exist; (ii) it might not be unique; and (iii) the network performance at equilibrium can be arbitrarily inefficient.

Our main result is to observe that a slight regulation on the network owners market solves all three issues above. Specially, if the authority could set appropriate caps (upper bounds) on the tolls (prices) operators can charge then: the game among the link operators has a unique strong Nash equilibrium and the users’ game results in a Wardrop equilibrium that achieves the optimal total delay. We call any price vector with these properties a great set of tolls. We then ask, can we compute great tolls that minimize total users’ payments? We show that this optimization problem reduces to a linear program in the case of single-commodity series-parallel networks. Starting from the same linear program, we obtain multiplicative approximation results for arbitrary networks with polynomial latencies of bounded degree, while in the single-commodity case we obtain a surprising bound, which only depends on the topology of the networkcase we obtain a surprising bound, which only depends on the topology of the network.


Evdokia Nikolova is an Assistant Professor in the Department of Electrical and Computer Engineering at the University of Texas at Austin, where she is a member of the Wireless Networking & Communications Group. She graduated with a BA in Applied Mathematics with Economics from Harvard University, MS in Mathematics from Cambridge University, U.K. and Ph.D. in Computer Science from MIT. Evdokia Nikolova’s research aims to improve the design and efficiency of complex systems (such as networks and electronic markets), by integrating stochastic, dynamic and economic analysis. Her recent work examines how human risk aversion transforms traditional computational models and solutions. One of her algorithms has been adapted in the MIT CarTel project for traffic-aware routing. She currently focuses on developing algorithms for risk mitigation in networks, with applications to transportation and energy. She is a recipient of an NSF CAREER award.

Red Hangover: Legacies of 20th Century Communism– Kristen Ghodsee

Kristen Ghodsee, Bowdoin | Wednesday, June 7, 2017


In Red Hangover Kristen Ghodsee examines the legacies of twentieth-century communism twenty-five years after the Berlin Wall fell. Ghodsee reflects on the lived experience of postsocialism and how many ordinary men and women across Eastern Europe suffered from the massive social and economic upheavals in their lives after 1989. Ghodsee shows how recent major crises—from the Russian annexation of Crimea and the Syrian Civil War to the rise of Islamic State and the influx of migrants in Europe—are linked to mistakes made after the collapse of the Eastern Bloc when fantasies about the triumph of free markets and liberal democracy blinded Western leaders to the human costs of “regime change.” Just as the communist ideal has become permanently tainted by its association with the worst excesses of twentieth-century Eastern European regimes, today the democratic ideal is increasingly sullied by its links to the ravages of neoliberalism. An accessible introduction to the history of European state socialism and postcommunism, Red Hangover reveals how the events of 1989 continue to shape the world today.


Kristen Ghodsee is a professor and the director of the Gender, Sexuality, and Women’s Studies Program at Bowdoin College. She is the author of seven books including, The Left Side of History; World War Two and the Unfulfilled Promise of Communism in Eastern Europe (Duke University Press, 2015) and Red Hangover: Legacies of 20th Century Communism, forthcoming with Duke University Press in October 2017. Ghodsee has held visiting fellowships at Harvard University, the Institute for Advanced Study in Princeton, and at the Freiburg Institute for Advanced Studies in Germany. In 2012, she won a Guggenheim Fellowship for her work in Anthropology and Cultural Studies.

Some Recent Advances in Scalable Optimization– Mladen Kolar

Mladen Kolar, U Chicago | Wednesday, May 3, 2017


In this talk, I will present two recent ideas that can help solve large scale optimization problems. In the first part, I will present a method for solving an ell-1 penalized linear and logistic regression problems where data are distributed across many machines. In such a scenario it is computationally expensive to communicate information between machines. Our proposed method requires a small number of rounds of communication to achieve the optimal error bound. Within each round, every machine only communicates a local gradient to the central machine and the central machine solves a ell-1 penalized shifter linear or logistic regression. In the second part, I will discuss usage of sketching as a way to solve linear and logistic regression problems with large sample size and many dimensions. This work is aimed at solving large scale optimization procedures on a single machine, while the extension to a distributed setting is work in progress.


Mladen Kolar is Assistant Professor of Econometrics and Statistics at the University of Chicago Booth School of Business. His research is focused on high-dimensional statistical methods, graphical models, varying-coefficient models and data mining, driven by the need to uncover interesting and scientifically meaningful structures from observational data. Particular applications arise in studies of dynamic regulatory networks and social media analysis. His research has appeared in several publications including the Journal of Machine Learning Research, Annals of Applied Statistics, and the Electronic Journal of Statistics. He also regularly presents his research at the top machine learning conferences, including Advances in Neural Information Processing Systems and the International Conference of Machine Learning.

Kolar was awarded a prestigious Facebook Fellowship in 2010 for his work on machine learning and network models. He spent a summer with Facebook’s ads optimization team working on a large scale system for click-through rate prediction. His other past research included work with INRIA Rocquencourt in Paris, France and Joint Research Center in Ispra, Italy.

Kolar earned his PhD in Machine Learning in 2013 from Carnegie Mellon University, as well as a diploma in Computer Engineering from the University of Zagreb. For his Ph.D. thesis work on “Uncovering Structure in High-Dimensions: Networks and Multi-task Learning Problems,” Kolar received from 2014 SIGKDD Dissertation Award honorable mention.

The ‘Mastery’ of the Swipe: Smartphones and Precarity in a Culture of Narcissism – Sharif Mowlabocus

Sharif Mowlabocus, University of Sussex | Wednesday, March 1, 2017 | Video


What do you do when you’re waiting for the bus?

Or waiting for a class to start?

Or waiting at the doctor’s office?

Or in line at the grocery store?

In this paper, I establish a dialogue between two discrete critical methodologies in order to consider the role of ‘distracted’ smartphone use within a socio-political context. By ‘distracted’ I am referring to the banal, everyday interactions we have with our smartphones throughout our day; the processes of swiping, tapping and gazing at our handheld devices, which occur dozens, if not hundreds of times a day, and which have taken on the appearance of a habit or social ‘tic’ (see also Caronia, 2005; Bittman et al. 2009).

Drawing on the work of Winnicott (1971) Lasch (1991), Silverstone (1993), Ribak, (2009) and Kullman (2010), I commute between psychoanalytic and political-economy methods in order to connect an analysis of distracted smartphone use to a broader discussion of social, political and economic precarity. Such an approach allows me to explore the relationship between the individual and society in order to identify how contemporary digital media practice is both a product of, and a response, to political, social and economic uncertainty.


Sharif Mowlabocus is a visiting researcher from the University of Sussex, UK, where he is a Senior Lecturer (Assoc. Prof.) in Digital Media. Sharif’s research is located at the intersection of LGBTQ studies and digital media studies. His work touches upon themes of digital embodiment, sexual representation and LGBTQ politics. While at MSRNE he will be working on a variety of projects that map onto these themes in different ways.

Schelling models in mathematics, social science, physics, and computer science – Richard Elwes

Richard Elwes, University of Leeds | Wednesday, December 7


In 1969, the economist Thomas Schelling devised some very simple theoretical models of racial segregation. His interest was in examples of the decoupling of “micromotives” from “macrobehaviour”, i.e. groups of agents who, by each acting according to their individual local preferences, cause a global effect desired by none of them. Schelling was unaware of the close resemblance of his models to others studied in depth since the early 20th century by statistical physicists. Later, similar constructions would appear within computer science as models of cascading phenomena on networks, and as neural nets. In this talk we will introduce Schelling models and survey some recent progress on their rigorous mathematical analysis.


Richard Elwes is a mathematician at University of Leeds (UK). His research background is in mathematical logic, but in recent years his interests have evolved to include the analysis of random processes arising in other sciences, and the interface between the two subjects. He is the author of five books on maths aimed at the general public, including Math 1001 and Chaotic Fishponds & Mirror Universes, and delivers regular talks and masterclasses to audiences of all types and levels.

Newtonian Regulation, Einsteinian Actors: The Magic Shoebox and the Contradictions of US Share Trading – Donald Mackenzie

Donald Mackenzie, University of Edinburgh | Wednesday, November 2, 2016


The ‘magic shoebox’ is a 38-mile coil of optical fibre in a computer datacentre in Secaucus, NJ known as ‘NY4’. For the last year, the world of professional share trading in the US has been convulsed by controversy about the shoebox.

The shoebox is deployed by a new stock exchange, IEX, which will be familiar to the readers of Michael Lewis’s 2014 bestseller Flash Boys. The purpose of the coil is to slow trading down, albeit by less than a thousandth of a second. The coil and the controversy surrounding it, MacKenzie will argue, have a much longer history than that sketched in Flash Boys. Their roots are in a decision in the late 1970s that was implicitly about how US share trading should be configured technologically, a decision whose long-time effects continue to shape how trading is regulated today.

Regulation abstracts away from the finite speed of signals and from the measurement of time and simultaneity, but in the ‘machine time’ of today’s high-frequency trading these issues matter. MacKenzie will sketch five features of today’s share trading that in part reflect the 1970s’ decision. The ‘shoebox’ is a response to these features, but in many ways epitomises the contradictions that afflict share trading rather than (as Lewis might have it) resolving them.

The talk will be based on an extensive, largely oral-historical study of the development of high-frequency trading (including the technologies and electronic trading venues that make it possible) and of its interactions with market regulation and – indirectly – with the US political system.


Donald MacKenzie is a professor of sociology at the University of Edinburgh. His current research is on the sociology of financial markets, especially the development of automated high-frequency trading (HFT), of the technologies and electronic trading venues that make it possible, and the interaction between HFT, regulation and the political system. His books include Inventing Accuracy: A Historical Sociology of Nuclear Missile Guidance (MIT Press, 1990) and An Engine, Not a Camera: How Financial Models Shape Markets (MIT Press, 2006). He writes regularly about financial markets in the London Review of Books.

A neurally-inspired model of habit and its empirical implications – Colin Camerer

Colin Camerer, Caltech | Wednesday, October 5, 2016 | Video


The busy human brain creates fast, low-cost habits when choices are frequent and are providing stable rewards. Using evidence from animal learning and cognitive neuroscience, we model a two-controller system in which habit and model-based choice coexist. The key inputs are reward prediction error (RPE) and the absolute magnitude of RPE. As the RPEs from a choice move toward zero, habits form. When the magnitude of averaged RPE exceeds a threshold, habits are overridden by model-based choice. The model contrasts with long-standing approach in economics (which relies on complementarity of consumption choice) and has several interesting properties that can be tested with behavioral and cognitive dataeconomics.


Professor Colin F. Camerer is the Robert Kirby Professor of Behavioral Finance and Economics at the California Institute of Technology (located in Pasadena, California), where he teaches cognitive psychology and economics. Professor Camerer earned a BA degree in quantitative studies from Johns Hopkins in 1977, and an MBA in finance (1979) and a Ph.D. in decision theory (1981, at age 22) from the University of Chicago Graduate School of Business. Before coming to Caltech in 1994, Camerer worked at the Kellogg, Wharton, and University of Chicago business schools. He studies both behavioral and experimental economics.

Prediction and Causation– Sendhil Mullainathan

Sendhil Mullainathan, Harvard | Wednesday, September 7, 2016


Machine learning tools provide powerful new tools for prediction. They are often criticized for being weak on causal inference. In this talk, I will describe how they can play a central role in areas that are thought to be causal in nature: policy making, theory testing and experimentation. I will illustrate using applications from crime, finance and behavioral economics.


Sendhil Mullainathan is Professor of Economics at Harvard and a MacArthur Fellow.  He is author with Eldar Shafir of Scarcity: How Having Too Little Means So Much and has done influential research on the way human psychology shapes economic decisions, especially in developing countries and among the poor in developed countries.  He has founded or led a number of institutions that have helped apply ideas from his research to improve the lives of the global poor (through the Jamil Poverty Action Lab and Ideas42 that he co-founded) and the poor in the United States (through serving as Chief Economist of the Consumer Financial Protection Bureau).  In recent years his research has increasingly turned, in a collaboration with Jon Kleinberg, to human-computer interaction and the ways that machine learning can both help us understand economic choice and improve the choices individuals make.  He has also been working on automated discover of patterns in health data that may help save lives.

Thin Spanning Trees and their Algorithmic Applications– Amin Saberi

Amin Saberi, Stanford | Wednesday, August 24, 2016


Motivated by Jaeger’s modular orientation conjecture, Goddyn asked the following question: A spanning tree of a graph G is called epsilon-thin if it contains at most an epsilon fraction of the edges of each cut in that graph. Is there a function f:(0,1)→ℤ such that every f(epsilon)-edge-connected graph has an epsilon-thin spanning tree? I will talk about our journey in search of such thin trees, their applications concerning traveling salesman problems, and unexpected connections to graph sparsification and the Kadison-Singer problem.


Amin Saberi is Associate Professor and 3COM faculty scholar in Stanford University. He received his B.Sc. from Sharif University of Technology and his Ph.D. from Georgia Institute of Technology in Computer Science. His research interests include algorithms, design and analysis of social networks, and applications. He is a recipient of the Terman Fellowship, Alfred Sloan Fellowship and a number of best paper awards.

Amin is also co-founder and chairman of NovoEd, a social learning environment designed in his research lab and used by universities such as Stanford, UC Berkeley, and University of Michigan as well as non-profit and for-profit institutions for offering courses to hundreds of thousands of learners around the world.

How Can Natural Language Processing Help Cure Cancer?– Regina Barzilay

Regina Barzilay, MIT | Wednesday, August 17, 2016


Cancer inflicts a heavy toll on our society. One out of seven women will be diagnosed with breast cancer during their lifetime, a fraction of them contributing to about 450,000 deaths annually worldwide. Despite billions of dollars invested in cancer research, our understanding of the disease, treatment, and prevention is still limited.

Majority of cancer research today takes place in biology and medicine. Computer science plays a minor supporting role in this process if at all. In this talk, I hope to convince you that NLP as a field has a chance to play a significant role in this battle. Indeed, free-form text remains the primary means by which physicians record their observations and clinical findings. Unfortunately, this rich source of textual information is severely underutilized by predictive models in oncology. Current models rely primarily only on structured data.

In the first part of my talk, I will describe a number of tasks where NLP-based models can make a difference in clinical practice. For example, these include improving models of disease progression, preventing over-treatment, and narrowing down to the cure. This part of the talk draws on active collaborations with oncologists from Massachusetts General Hospital (MGH).

In the second part of the talk, I will push beyond standard tools, introducing new functionalities and avoiding annotation-hungry training paradigms ill-suited for clinical practice. In particular, I will focus on interpretable neural models that provide rationales underlying their predictions, and semi-supervised methods for information extraction.


Regina Barzilay is a professor in the Department of Electrical Engineering and Computer Science and a member of the Computer Science and Artificial Intelligence Laboratory at the Massachusetts Institute of Technology. Her research interests are in natural language processing. She is a recipient of various awards including of the NSF Career Award, the MIT Technology Review TR-35 Award, Microsoft Faculty Fellowship and several Best Paper Awards at NAACL and ACL. She received her Ph.D. in Computer Science from Columbia University, and spent a year as a postdoc at Cornell University.

Learning and Equilibrium in Games – Drew Fudenberg

Drew Fudenberg, Harvard | Wednesday, August 3, 2016 | Video


When and why will observed play in a game approximate an equilibrium, and what sort of equilibria will persist? To understand this, we study the long-run outcomes of rational non-equilibrium learning. In one-shot simultaneous-move games, steady states of such processes must be Nash equilibria, but this is not true in   extensive- form games, where mistaken beliefs about opponents’ play and non-Nash outcomes can persist due to the tradeoff between exploration and exploitation.   When players are patient, learning leads players to have the correct beliefs about the path of play and so to a subset of the Nash equilibria. Ongoing research analyzes this subset for the class of signalling games, which are known to have many implausible Nash equilibria.


Drew Fudenberg is the Paul A. Samuelson Professor of Economics at MIT.  He is best known for his work on game theory, which ranges from foundational work on learning and equilibrium to the study of  particular games used to study e.g. reciprocal altruism, reputation effects,   and competition between firms. More recently he has also worked on topics in behavioral economics and decision theory. He is the author of four books, including a leading game theory text.  He received an A.B. in applied mathematics from Harvard College in 1978, and a Ph.D. in economics from MIT in 1981, followed by faculty appointments at UC Berkeley, MIT, and Harvard.  He is a Fellow of the Econometric Society, and will be its President in 2017. He is a member of both the National Academy of Sciences and the American Academy of Arts and Sciences. He is also a past editor of Econometrica and a co-founder of the open access journal Theoretical Economics.

Black + Twitter: A Cultural Informatics Approach – Andre Brock

Andre Brock, U Michigan | Wednesday, July 27, 2016 | Video


Chris Sacca, activist investor, recently argued that Twitter IS Black Twitter.  African American usage of the service often dominates user metrics in the United States, despite their minority demographic status among computer users.  This talk unpacks Black Twitter use from two perspectives: analysis of the interface and associated practice alongside discourse analysis of Twitter’s utility and audience.  Using examples of Black Twitter practice, I offer that Twitter’s feature set and ubiquity map closely onto Black discursive identity.  Thus, Twitter’s outsized function  as mechanism for cultural critique  and political activism can be understood as the awakening of Black digital practice and an abridging of a digital divide.


André Brock is an Assistant Professor of Communication Studies at the University of Michigan. Brock is one of the preeminent scholars of Black Cyberculture. His work bridges Science and Technology Studies and Critical Discourse Analysis, showing how the communicative affordances of online media align with those of Black communication practices. Through December 2016, he is a Visiting Researcher with the Social Media Collective at Microsoft Research New England.

Engaging Patients In Their Own Health: Will Predictive Analytics Help To Sharpen Blunt Instruments? – Niteesh Choudhry

Niteesh Choudhry, Harvard Medical School | Wednesday, July 20, 2016


Many highly effective health interventions are never widely adopted into routine care. For example, each year only 40% of adults receive a flu vaccine and only half of patients who have had a heart attack continue to take their cardiac medications over the long-term. Overcoming these gaps in health care implementation requires patients and providers to be more actively involved in healthcare delivery. Engagement techniques that rely on strategies from behavioral economics, marketing, cognitive psychology, information technology and other related disciplines have shown promise, although even the best of these interventions have changed behavior to only a modest extent. This may result from either the interventions themselves having limited effectiveness or their not being optimally targeted to those who may benefit the most. In this lecture, I will use the example of medication non-adherence, an exceptionally common public health problem that annually accounts for hundreds of billions of dollars of potentially avoidable health spending in the U.S. alone, to describe how predictive analytics is being used to refine patient engagement interventions. I will review both what appears to be possible today (as in identifying who is likely to be non-adherent in the future and when they will exhibit this behavior) and describe what else needs to be developed to fully capture an individual’s behavioral response phenotype.


Niteesh K. Choudhry, MD, PhD, is an Associate Professor at Harvard Medical School and Executive Director of the Center for Healthcare Delivery Sciences ( at Brigham and Women’s Hospital, where he also Associate Physician in the Division of Pharmacoepidemiology and Pharmacoeconomics. Much of Dr. Choudhry’s recent work has dealt with the design and implementation of large simple clinical trials embedded in real-world health systems. He was the principal investigator of the Post-MI Free Rx and Event and Economic Evaluation (MI FREEE) trial, on the basis of which Aetna has changed their benefits to waive medication copayments for post-MI secondary prevention medications. He is also the principal investigator of several other large pragmatic clinical trials designed to engage patients in improving their own health care. These include the Randomized Evaluation to Measure Improvements in Nonadherence from low-cost Devices (REMIND) trials, the NHLBI-funded Study of a Telepharmacy Intervention for Chronic disease to Improve Treatment adherence (STIC 2 IT), and the Targeted Adherence intervention to Reach Glycemic control with Insulin Therapy for Diabetes patients (TARGIT – Diabetes) study and the ENhancing outcomes through Goal Assessment and Generating Engagement in Diabetes Mellitus (ENGAGE-DM) trial. He is also the Co-Principal Investigator of the Mail Outreach To Increase Vaccine Acceptance Through Engagement (MOTIVATE) trial, conducted in partnership with the White House Social and Behavioral Sciences Team and the Center for Medicare and Medicaid Services; it seeks to increase rates of influenza vaccination among Medicare beneficiaries.Dr. Choudhry attended McGill University, received his M.D. and completed his residency training in Internal Medicine at the University of Toronto and then served as Chief Medical Resident for the Toronto General and Toronto Western Hospitals. He earned his Ph.D. in Health Policy from Harvard University. He has published over 190 scientific papers in leading medical and policy journals and has won awards from AcademyHealth, the Society of General Internal Medicine, the International Society of Pharmacoeconomics and Outcomes Research, and the National Institute of Health Care Management for his research. His work is supported by both public and private funders including the National Heart, Lung, and Blood Institute, the Agency for Healthcare Quality and Research, CVS Caremark, Aetna, the Robert Wood Johnson Foundation, the Commonwealth Fund, the Arnold Foundation, Merck, Sanofi, AstraZeneca and the Pharmaceutical Research and Manufacturers of America.Dr. Choudhry practices inpatient general internal/hospital medicine and has won numerous awards for teaching excellence.

Personal Control of Data – Butler Lampson

Butler Lampson, MSR New England | Wednesday, July 13, 2016 | Video


People around the world are concerned that more and more of their personal data is on the Internet, where it’s easy to find, copy, and link up with other data. Data about people’s presence and actions in the physical world (from cameras, microphones, and other sensors) soon will be just as important as data that is born digital. What people most often want is a sense of control over their data (even if they don’t exercise this control very often). Control means that you can tell who has your data, limit what they can do with it, and change your mind about the limits. Many people feel that this control is a fundamental human right (thinking of personal data as an extension of the self), or an essential part of your property rights to your data.Regulators are starting to respond to these concerns. Because societies around the world have different cultural norms and governments have different priorities, there will not be a single worldwide regulatory regime. However, it does seem possible to have a single set of basic technical mechanisms that support regulation, based on the idea of requiring data holders to respect the current policy of data subjects about how their data is used.


Butler is a Technical Fellow at Microsoft and an Adjunct Professor at MIT.  He has worked on computer architecture, local area networks, raster printers, page description languages, operating systems, remote procedure call, programming languages and their semantics, programming in the large, fault-tolerant computing, transaction processing, computer security, WYSIWYG editors, and tablet computers.  He was one of the designers of the SDS 940 time-sharing system, the Alto personal distributed computing system, the Xerox 9700 laser printer, two-phase commit protocols, the Autonet LAN, the SPKI system for network security, the Microsoft Tablet PC software, the Microsoft Palladium high-assurance stack, and several programming languages. He received the ACM Software Systems Award in 1984 for his work on the Alto, the IEEE Computer Pioneer award in 1996 and von Neumann Medal in 2001, the Turing Award in 1992, and the NAE’s Draper Prize in 2004.

Incentive Alignment for Machine Learning – Yiling Chen

Yiling Chen, Harvard | Wednesday, June 29, 2016


We are blessed with unprecedented abilities to connect with people all over the world: buying and selling products, sharing information and experiences, asking and answering questions, collaborating on projects, borrowing and lending money, and exchanging excess resources. These activities result in rich data that scientists can use to understand human social behavior, generate accurate predictions, and make policy recommendations. Machine learning traditionally take such data as given, often treating them as independent samples drawn from some underlying true distribution. However, such data are possessed or generated by (potentially strategic) people in the context of specific interaction rules. Hence, what data become available depends on the interaction rules. In this talk, I argue that a holistic view that jointly considers data acquisition and inference and learning is important for machine learning. As an example, I will present a project on incentivizing strategic agents to generate high-quality data for the purpose of regression.


Yiling Chen is the Gordon McKay Professor of Computer Science at Harvard University. She received her Ph.D. in Information Sciences and Technology from the Pennsylvania State University. Prior to working at Harvard, she spent two years at Yahoo! Research in New York City. Her current research focuses on topics in the intersection of computer science and economics. Her awards include an ACM EC Outstanding Paper Award, an AAMAS Best Paper Award, an NSF Career award and The Penn State Alumni Association Early Career Award, and she was selected by IEEE Intelligent Systems as one of “AI’s 10 to Watch” in 2011.

Rigorous Foundations for Privacy in Statistical Databases – Adam Smith

Adam Smith, Penn State | Wednesday, June 22, 2016


Consider an agency holding a large database of sensitive personal information — medical records, census survey answers, web search records, or genetic data, for example. The agency would like to discover and publicly release global characteristics of the data (say, to inform policy or business decisions) while protecting the privacy of individuals’ records. This problem is known variously as “statistical disclosure control”, “privacy-preserving data mining” or “private data analysis”. I will begin by discussing what makes this problem difficult, and exhibit some of the nontrivial issues that plague simple attempts at anonymization and aggregation. Motivated by this, I will present differential privacy, a rigorous definition of privacy in statistical databases that has received significant attention. I’ll explain some recent results on the design of differentially private algorithms, as well as the application of these ideas in contexts with no (previously) apparent connection to privacy.


Adam Smith is a professor of Computer Science and Engineering at Penn State. His research interests lie in data privacy and cryptography and their connections to information theory, statistical learning and quantum computing. He received his Ph.D. from MIT in 2004 and was subsequently a visiting scholar at the Weizmann Institute of Science and UCLA and a visiting professor at Boston University and Harvard. He received a 2009 Presidential Early Career Award for Scientists and Engineers (PECASE) and the 2016 Theory of Cryptography Test of Time Award (with Dwork, McSherry and Nissim).

Incentive Auctions and Spectrum Repacking: A Case Study for Deep Optimization – Kevin Leyton-Brown

Kevin Leyton-Brown, U British Columbia | Wednesday, June 15, 2016


This talk will discuss the FCC’s “incentive auction”–currently underway!–which proposes to give television broadcasters an opportunity to sell their broadcast rights, to repack remaining broadcasters into a smaller block of spectrum, and to resell the freed airwaves to telecom companies. The stakes for this auction are huge–projected tens of billions of dollars in revenue for the government–justifying the design of a special-purpose descending-price auction mechanism. An inner-loop problem in this mechanism is determining whether a given set of broadcasters can be repacked into a smaller block of spectrum while respecting radio interference constraints. This is an instance of a (worst-case intractable) graph coloring problem; however, stations’ broadcast locations and interference constraints are all known in advance. Early efforts to solve this problem considered hand-crafted mixed-integer programming formulations, but were unable to reliably solve realistic, national-scale problem instances. We advocate instead for a “deep optimization” approach that applies abundant offline computation to tailor an algorithm to the problem at hand. In particular, we leveraged automatic algorithm configuration and algorithm portfolio techniques, alongside constraint graph decomposition; novel caching mechanisms that allow reuse of partial solutions from related, solved problems; and the marriage of local-search and complete SAT solvers. We show that our approach solves virtually all of a set of problems derived from auction simulations within the short time budget required in practice.


Kevin Leyton-Brown is a professor of computer science at the University of British Columbia. He studies the intersection of computer science and microeconomics, addressing computational problems in economic contexts and incentive issues in multiagent systems. He also applies machine learning to the automated design and analysis of algorithms for solving hard computational problems. Lately, he has been involved in designing an algorithm to clear the FCC’s upcoming “incentive auction” for radio spectrum; applying deep learning to model human behavior in games and discrete choice settings; building an online market for agricultural commodities in Uganda (“Kudu”); and building a system for TA-supported student peer grading (“Mechanical TA”) and analyzing its game theoretic properties. Kevin received his PhD from Stanford University. He is the recipient of UBC’s 2015 Charles A. McDowell Award for Excellence in Research, a 2014 NSERC E.W.R. Steacie Memorial Fellowship, and a 2013 Outstanding Young Computer Science Researcher Prize from the Canadian Association of Computer Science.

Decision making at scale: Algorithms, Mechanisms, and Platforms – Ashish Goel

Ashish Goel, Stanford | Wednesday, May 25, 2016


YouTube competes with Hollywood as an entertainment channel, and also supplements Hollywood by acting as a distribution mechanism. Twitter has a similar relationship to news media, and Coursera to Universities. But there are no online alternatives for making democratic decisions at large scale as a society. In this talk, we will describe two algorithmic approaches towards large scale decision making that we are exploring. We will also describe our experience with helping implement participatory budgeting in close to two dozen cities and municipalities, including Cambridge, MA, and briefly comment on issues of fairness.

  1. Knapsack voting and participatory budgeting: All budget problems are knapsack problems at their heart, since the goal is to pack the largest amount of societal value into a budget. This naturally leads to “knapsack voting” where each voter solves a knapsack problem, or comparison-based voting where each voter compares pairs of projects in terms of benefit-per-dollar. We analyze natural aggregation algorithms for these mechanisms, and natural utility models for voters, and show that knapsack voting is strategy-proof under these models.
  2. Triadic consensus: Here, we divide individuals into small groups (say groups of three) and ask them to come to consensus; the results of the triadic deliberations in each round form the input to the next round. We show that this method is efficient and incentivizes truth-telling in fairly general settings, whereas no pair-wise deliberation process can have the same properties.

This is joint work with Tanja Aitamurto, Brandon Fain, Anilesh Krishnaswamy, David Lee, Kamesh Munagala, and Sukolsak Sakshuwong.


Ashish Goel is a Professor of Management Science and Engineering and (by courtesy) Computer Science at Stanford University, and a member of Stanford’s Institute for Computational and Mathematical Engineering. He received his PhD in Computer Science from Stanford in 1999, and was an Assistant Professor of Computer Science at the University of Southern California from 1999 to 2002. His research interests lie in the design, analysis, and applications of algorithms; current application areas of interest include social networks, participatory democracy, Internet commerce, and large scale data processing. Professor Goel is a recipient of an Alfred P. Sloan faculty fellowship (2004-06), a Terman faculty fellowship from Stanford, an NSF Career Award (2002-07), and a Rajeev Motwani mentorship award (2010). He was a co-author on the paper that won the best paper award at WWW 2009, and an Edelman Laureate in 2014.

How Spreadsheets Shape Organizational Life: A Case Study in the Materialities of Information – Paul Dourish

Paul Dourish, UC Irvine | Wednesday, May 18, 2016 | Video


Material manifestations of digital representations — as bits of a disk, signals in a wire, or images on a screen — shape, constrain, and enable various forms of human and social action. Based on a book in progress that draws on a range of examples including network protocols and database architectures, this talk will focus on example of spreadsheets to make this argument. A number of recent studies have examined the role that Powerpoint plays in organizational life. Organizational scholars like Wanda Orlikowski and Joanne Yates have looked at the Powerpoint presentation as a particular genre of organizational practice; critics like Edwards Tufte have bemoaned the dumbing-down of powerpoint-driven communication. In ethnographic work of large-scale science, my colleagues and I were struck by a related phenomenon, which is the prevalence of spreadsheets, not just as a document format, but as something that gets incorporated into meetings. This talk will show how the ways spreadsheets are designed — the constraints and shapes they offer and require — structure talk and action in particular ways that get organizational work done.


Paul Dourish is Professor of Informatics at UC Irvine, with courtesy appointments in Computer Science and Anthropology; he also has visiting appointments at the University of Melbourne and with Comparative Media Studies at MIT. His research lies primarily in human-computer interaction, software studies, science and technology studies, and cultural studies of digital media. Before joining the faculty at Irvine, he was Senior Research Scientist in the Computer Science Laboratory at Xerox PARC. His current book project, for MIT Press, explores the representational materialities of digital information through a range of case studies including Internet routing algorithms, databases, and spreadsheets. He is a Fellow of the ACM, a member of the ACM SIGCHI Academy, and has received the AMIA Diana Forsythe Award and the CSCW “Lasting Impact” award.

Genetic Screens with CRISPR: A New Hope in Functional Genomics – John Doench

John Doench, Broad Institute | Wednesday, May 4, 2016 | Video


Functional genomics attempts to understand the genome by disrupting the flow of information from DNA to RNA to protein and then observing how the cell or organism changes in response. Both RNAi and CRISPR technologies are simply hacks of systems that originally evolved to silence viruses, reprogrammed to target genes we’re interested in studying, as decoding the function of genes is a critical step towards understanding how gene dysfunction leads to disease. Here we will discuss the development and optimization of CRISPR technology for genome-wide genetic screens and its application to multiple biological problems.


John Doench is the Associate Director of the Genetic Perturbation Platform at the Broad Institute. He develops and applies the latest approaches in functional genomics, including RNAi, ORF, and CRISPR technologies, to understand the function of genes and how gene dysfunction leads to disease. John collaborates with researchers across the community to develop faithful biological models and execute genetic screens. Prior to joining the Broad in 2009, John did his postdoctoral work at Harvard Medical School, received his PhD from the biology department at MIT, and majored in history at Hamilton College. John lives in Jamaica Plain, MA with his wife and daughter, where he enjoys coaching soccer, cheering on the Red Sox and Patriots, playing volleyball, running, and avoiding imminent death while navigating the streets of Boston on a bicycle.

Collective Graph Identification – Lise Getoor

Lise Getoor, UC Santa Cruz | Wednesday, April 13, 2016


Graph data (e.g., communication data, financial transaction networks, data describing biological systems, collaboration networks, the Web, etc.) is ubiquitous. While this observational data is useful, it is usually noisy, often only partially observed, and only hints at the actual underlying social, scientific or technological structures that give rise to the interactions. For example, an email communication network provides useful insight, but is not the same as the “real” social network among individuals. In this talk, I introduce the problem of graph identification, i.e., the discovery of the true graph structure underlying an observed network. This involves inferring the nodes, edges, and node labels of a hidden graph based on evidence provided by the observed graph. I show how this can be cast as a collective probabilistic inference task and describe a scalable approach to solving this problem.


Lise Getoor is a professor in the Computer Science Department at the University of California, Santa Cruz. Her research areas include machine learning, data integration and reasoning under uncertainty, with an emphasis on graph and network data. She has over 200 publications and extensive experience with machine learning and probabilistic modeling methods for graph and network data. She is a Fellow of the Association for Artificial Intelligence, an elected board member of the International Machine Learning Society, serves on the board of the Computing Research Association (CRA), and was co-chair for ICML 2011. She is a recipient of an NSF Career Award and ten best paper and best student paper awards. She received her PhD from Stanford University in 2001, her MS from UC Berkeley, and her BS from UC Santa Barbara, and was a professor in the Computer Science Department at the University of Maryland, College Park from 2001-2013.

Reframing the Financial Inclusion Debate: Evidence from an Up-Close View of Check Cashers and Payday Lenders – Lisa Servon

Lisa Servon, New School | Wednesday, April 6, 2016 | Video


What do a Mexican immigrant living in the South Bronx, a twenty-something graduate student, and a telemarketer in Dallas have in common? All three are victims of our dysfunctional mainstream bank and credit system. As banks have grown larger and focused less on serving ordinary consumers, many have begun to get their financial needs meet from alternative financial services providers like check cashers and predatory lenders. Although these businesses are labeled as predatory and sleazy, their customers find that they offer three things banks no longer provide: less expensive products and services, greater transparency, and better service. At a time when 57 percent of Americans are struggling financially, and trust in banks is at an all-time low, it’s imperative that we understand how we got here, and what we can do to make financial health a reality for all Americans.


Lisa J. Servon is Professor and former dean at the Milano School of International Affairs, Management, and Urban Policy at The New School. She is currently a Scholar at the Russell Sage Foundation. Professor Servon holds a PhD in Urban Planning from the University of California, Berkeley, an MA in the History of Art from the University of Pennsylvania and a BA from Bryn Mawr College. She teaches and conducts research in the areas of urban poverty, community development, economic development, and issues of gender and race. Her current research focuses on the alternative financial services industry. Her book, The Unbanking of American: How the New Middle Class Survives, will be published by Houghton Mifflin Harcourt in 2017. She spent 2004-2005 as Senior Research Fellow at the New America Foundation in Washington, DC. Servon is the author or editor of numerous journal articles and four books: Bridging the Digital Divide: Technology, Community, and Public Policy (Blackwell 2002), and Bootstrap Capital: Microenterprises and the American Poor (Brookings 1999), Gender and Planning: A Reader (With Susan Fainstein, Rutgers University Press 2005), and Otra Vida es Posible: Practicas Economicas Alternativas Durante la Crisis (With Manuel Castells, Joana Conill, Amalia Cardenas and Sviatlana Hlebik. UOC Press 2012).

Security Games: Key Algorithmic Principles, Deployed Applications and Research Challenges – Milind Tambe

Milind Tambe, USC | Wednesday, March 16, 2016 | Video


Security is a critical concern around the world, whether it is the challenge of protecting ports, airports and other critical infrastructure, protecting endangered wildlife, forests and fisheries, suppressing urban crime or security in cyberspace. Unfortunately, limited security resources prevent full security coverage at all times; instead, we must optimize the use of limited security resources. To that end, our “security games” framework — based on computational game theory, while also incorporating elements of human behavior modeling, AI planning under uncertainty and machine learning — has led to building and deployment of decision aids for security agencies in the US and around the world. These decision aids are in use by agencies such as the US Coast Guard, the Federal Air Marshals Service and by various police agencies at university campuses, airports and metro trains. Moreover, recent work on “green security games” has led our decision aids to begin assisting NGOs in protection of wildlife; and “opportunistic crime security games” have focused on suppressing urban crime. I will discuss our use-inspired research in security games that is leading to new research challenges, including algorithms for scaling up security games as well as for handling significant adversarial uncertainty and learning models of human adversary behaviors.

Joint work with a number of current and former PhD students, postdocs all listed here.


Milind Tambe is Helen N. and Emmett H. Jones Professor in Engineering at the University of Southern California(USC). He is a fellow of AAAI and ACM, as well as recipient of the ACM/SIGART Autonomous Agents Research Award, Christopher Columbus Fellowship Foundation Homeland security award, INFORMS Wagner prize for excellence in Operations Research practice, Rist Prize of the Military Operations Research Society, IBM Faculty Award, Okawa foundation faculty research award, RoboCup scientific challenge award, and other local awards such as the Orange County Engineering Council Outstanding Project Achievement Award, USC Associates award for creativity in research and USC Viterbi use-inspired research award. Prof. Tambe has contributed several foundational papers in AI in areas such as multiagent teamwork, distributed constraint optimization (DCOP) and security games. For this research, he has received the influential paper award and a dozen best paper awards at conferences such as AAMAS, IJCAI, IAAI and IVA. In addition, Prof. Tambe pioneering real-world deployments of ”security games” has led him and his team to receive the US Coast Guard Meritorious Team Commendation from the Commandant, US Coast Guard First District’s Operational Excellence Award, Certificate of Appreciation from the US Federal Air Marshals Service and special commendation given by LA Airport police from the city of Los Angeles. For his teaching and service, Prof. Tambe has received the USC Steven B. Sample Teaching and Mentoring award and the ACM recognition of service award. He has also co-founded a company based on his research, ARMORWAY, where he serves as the director of research. Prof. Tambe received his Ph.D. from the School of Computer Science at Carnegie Mellon University.

Paradoxes of Openness and Distinction in the Sharing Economy – Juliet Schor

Juliet Schor, Boston College | Wednesday, December 16, 2015 | Video


Since the 1980s, Pierre Bourdieu’s influence in sociology has increased markedly, including on the study of consumption and economic life (Sallaz and Zavisca 2007). Bourdieu’s formulation of multiple types of capital (economic, cultural and social) and their role in producing and reproducing durable inequality has been highly productive in a variety of contexts. However, while scholars have examined practices of distinction, the structure of particular fields, and the role of specific capitals in social reproduction, there has been less attention to economic exchanges at a micro, interactional level (King 2000). In this paper, we use a Bourdieusian approach to study new kinds of exchanges in the “sharing economy” and the ways in which distinction and inequality operate within them. To do this, we extend Bourdieu by bringing in conceptual tools from relational economic sociology. This literature, pioneered by Viviana Zelizer (2010, 2005b, 2012), emphasizes the importance of meaning, the role of culture in structuring economic activity, and the idea that economic exchanges require ongoing interpersonal negotiations. We use relational analysis to study how people deploy, convert, and use their capital. In particular, we show how cultural capital is used to establish superior position in the context of various types of exchanges. Thus, our contribution is an investigation into how Bourdieusian inequality is reproduced via interpersonal relations in the context of exchange.


Juliet Schor is Professor of Sociology at Boston College. She is also a member of the MacArthur Foundation Connected Learning Research Network. Schor’s research focuses on issues of time use, consumption and environmental sustainability. A graduate of Wesleyan University, Schor received her Ph.D. in economics at the University of Massachusetts. Before joining Boston College, she taught at Harvard University for 17 years, in the Department of Economics and the Committee on Degrees in Women’s Studies. In 2014 Schor received the American Sociological Association’s award for Public Understanding of Sociology. She also served as the Matina S. Horner Distinguished Visiting Professor at the Radcliffe Institute at Harvard University.

Schor’s most recent books are Sustainable Lifestyles and the Quest for Plenitude: Case Studies of the New Economy (Yale University Press, 2014) which she co-edited with Craig Thompson, and True Wealth: How and Why Millions of Americans are Creating a Time-Rich, Ecologically Light, Small-Scale, High-Satisfaction Economy (2011 by The Penguin Press, previously published as Plenitude. As part of her work with the MacArthur Foundation, Schor is currently researching the “connected economy,” via a series of case studies of sharing platforms and their participants. She is also studying the relation between working hours, carbon emissions and economic growth.

Schor’s previous books include the national best-seller The Overworked American: The Unexpected Decline of Leisure (Basic Books, 1992) and The Overspent American: Why We Want What We Don’t Need (Basic Books, 1998). She appears frequently on national and international media, and profiles on her and her work have appeared in scores of magazines and newspapers, including The New York Times, Wall Street Journal, Newsweek, and People magazine. She has appeared on 60 Minutes, the Today Show, Good Morning America, The Early Show on CBS, numerous stories on network news, as well as many other television and radio news programs.

The Strange Logic of Galton-Watson Trees – Joel Spencer

Joel Spencer, NYU | Wednesday, December 2, 2015 | Video


The Galton-Watson tree is a basic demographic model. The classic equation for a Galton-Watson tree being infinite has two solutions, only one of which is “correct.” What about other properties. (Example: Some node has precisely two children.) We show that when the property is what is called first order than there is a unique solution to the corresponding equation. We consider “tree automata” and the situation for monadic second order properties.


Joel Spencer is a Professor of Mathematics and Computer Science at the Courant Institute, New York. His work is at the fecund intersection of Probability, Discrete Math and Logic, with a strong asymptotic flavor. He is a disciple of Paul Erdos. A new edition of His book (with Noga Alon) The Probabilistic Method will be published in December.

The Rise of the Sharing Economy: Estimating the Impact of Airbnb on the Hotel Industry – Giorgos Zervas

Giorgos Zervas, BU | Wednesday, November 18, 2015


A number of decentralized peer-to-peer markets, now colloquially known as the sharing economy, have emerged as alternative suppliers of goods and services traditionally provided by long-established industries. A central question surrounding the sharing economy regards its long-term impact: will peer-to-peer platforms materialize as viable mainstream alternatives to traditional providers, or will they languish as niche markets? In this paper, we study Airbnb, a sharing economy pioneer offering short-term accommodation. Combining data from Airbnb and the Texas hotel industry, we estimate the impact of Airbnb’s entry into the Texas market on hotel room revenue, and study the market response of hotels. To identify Airbnb’s causal impact on hotel room revenue, we use a difference-in-differences empirical strategy that exploits the significant spatiotemporal variation in the patterns of Airbnb adoption across city-level markets. We estimate that in Austin, where Airbnb supply is highest, the impact on hotel revenue is roughly 8-10%. We find that Airbnb’s impact is non-uniformly distributed, with lower-priced hotels, and hotels not catering to business travel being the most affected segments. Finally, we find that affected hotels have responded by reducing prices, an impact that benefits all consumers, not just participants in the sharing economy. Our work provides empirical evidence that the sharing economy is making inroads by successfully competing with, and acquiring market share from, incumbent firms.


Georgios Zervas is an assistant professor of Marketing at Questrom School of Business at Boston University. Before joining BU in 2013 he was a Simons postdoctoral fellow at Yale, and an affiliate at the Center for Research on Computation and Society at Harvard. He received his PhD in Computer Science in 2011 from Boston University. He is broadly interested in understanding the strategic interactions of firms and consumers participating in internet markets using large-scale data collection and econometric analysis.

Comics and Stuff: An Introduction – Henry Jenkins

Henry Jenkins, USC | Wednesday, November 11, 2015 | Video


The status and nature of comics are under transition, as comics move from a disposable medium to one which is perceived as having enduring value. The emergence of the so-called “graphic novel” represents a shift in how comics are published, in terms of what kind of cultural status they command, in terms of who reads and writes them, in terms of how people access them, and in terms of what kinds of stories they tell. Comics artists and readers have historically been collectors who sorted through this “trash” medium to decide what should be kept and discarded. And today’s graphic novels often telling “collecting stories,” that is, stories by, for and about collectors, using their protagonist’s relationships with material objects as a means of sorting through their own relationship to this evolving medium. This talk will draw insights from contemporary work in cultural anthropology, literary criticism, and art history that speaks about “stuff,” “objects,” “things,” to think about the ways contemporary comics represent our relationship to the material world and through this, reflect on our relationship to issues of memory, nostalgia, and history.


Henry Jenkins is the Provost’s Professor of Communication, Journalism, Cinematic Art, and Education at the University of Southern California. He is the author or editor of 17 books on various aspects of media change and popular culture, including Textual Poachers: Television Fans and Participatory Culture, Convergence Culture: Where Old and New Media Collide, Spreadable Media: Creating Meaning and Value in a Networked Society, Participatory Culture in a Networked Era, and By Any Media Necessary: The New Activism of American Youth.

Recovering usable hidden structure using exploratory data analyses on genomic data – Barbara Engelhardt

Barbara Engelhardt, Princeton | Wednesday, November 4, 2015


Methods for exploratory data analysis have been the recent focus of much attention in `big data’ applications because of their ability to quickly allow the user to explore structure in the underlying data in a controlled and interpretable way. In genomics, latent factor models are commonly used to identify population substructure, identify gene clusters, and control noise in large data sets. In this talk I will describe a series of statistical models for exploratory data analysis to illustrate the structure that they are able to identify in large genomic data sets. I will consider several downstream uses for the recovered latent structure: understanding technical noise in the data, developing undirected networks from the recovered structure, and using this latent structure to study genomic differences among people.


Barbara Engelhardt is an assistant professor in the Computer Science Department and the Center for Statistics and Machine Learning at Princeton University. Prior to that, she was at Duke University as an assistant professor in Biostatistics and Bioinformatics and Statistical Sciences. She graduated from Stanford University and received her Ph.D. from the University of California, Berkeley, advised by Professor Michael Jordan. She did postdoctoral research at the University of Chicago, working with Professor Matthew Stephens. Interspersed among her academic experiences, she spent two years working at the Jet Propulsion Laboratory, a summer at Google Research, and a year at 23andMe, a personal genomics company. Professor Engelhardt received an NSF Graduate Research Fellowship, the Google Anita Borg Memorial Scholarship, and the Walter M. Fitch Prize from the Society for Molecular Biology and Evolution. She also received the NIH NHGRI K99/R00 Pathway to Independence Award. Professor Engelhardt is currently a PI on the Genotype-Tissue Expression (GTEx) Consortium. Her research interests involve statistical models and methods for analysis of high-dimensional data, with a goal of understanding the underlying biological mechanisms of complex phenotypes and human diseases.

Improving Urban Public Education: Lessons from Charter Schools – Parag Pathak

Parag Pathak, MIT | Wednesday, October 28, 2015 | Video


Charter schools represent one of the fastest growing, yet controversial innovations in education reform. In this talk, I will review several papers measuring urban school performance from a series of papers using data from Boston, New York City, Denver, and New Orleans from MIT’s School Effectiveness and Inequality Initiative. In addition to discussing the broader debates on sources of achievement gaps, I will also briefly touch upon some new methodological issues emerging from this work.


Parag A. Pathak is a Professor of Economics at MIT, founding co-director of the NBER Working Group on Market Design, and founder of MIT’s School Effectiveness and Inequality Initiative (SEII), a laboratory focused on education, human capital, and the income distribution. His work on market design and education was recognized with a Presidential Early Career Award for Scientists and Engineers, an Alfred P. Sloan Fellowship, the Shapley Lectureship, and the 2016 Social Choice and Welfare Prize. More than a million students have been assigned to school in choice systems he has helped to design in Boston, Chicago, Denver, New Orleans, New York, and Washington DC.

The Contextual Bandits Problem: A New, Fast, and Simple Algorithm – Robert Schapire

Robert Schapire, MSR-NYC | Wednesday, October 14, 2015 | Video


We study the general problem of how to learn through experience to make intelligent decisions. In this setting, called the contextual bandits problem, the learner must repeatedly decide which action to take in response to an observed context, and is then permitted to observe the received reward, but only for the chosen action. The goal is to learn through experience to behave nearly as well as the best policy (or decision rule) in some possibly very large and rich space of candidate policies. Previous approaches to this problem were all highly inefficient and often extremely complicated. In this work, we present a new, fast, and simple algorithm that learns to behave as well as the best policy at a rate that is (almost) statistically optimal. Our approach assumes access to a kind of oracle for classification learning problems which can be used to select policies; in practice, most off-the-shelf classification algorithms could be used for this purpose. Our algorithm makes very modest use of the oracle, which it calls far less than once per round, on average, a huge improvement over previous methods. These properties suggest this may be the most practical contextual bandits algorithm among all existing approaches that are provably effective for general policy classes.

This is joint work with Alekh Agarwal, Daniel Hsu, Satyen Kale, John Langford and Lihong Li.


Robert Schapire is a Principal Researcher at Microsoft Research in New York City. He received his PhD from MIT in 1991. After a short post-doc at Harvard, he joined the technical staff at AT&T Bell Laboratories (later, AT&T Labs) in 1991. In 2002, he became a Professor of Computer Science at Princeton University where he was later named the David M. Siegel ’83 Professor in Computer Science. He joined Microsoft Research in 2014. His awards include the 1991 ACM Doctoral Dissertation Award, the 2003 Gödel Prize, and the 2004 Kanelakkis Theory and Practice Award (both of the last two with Yoav Freund). He is a fellow of the AAAI, and a member of the National Academy of Engineering. His main research interest is in theoretical and applied machine learning, with particular focus on boosting, online learning, game theory, and maximum entropy.

An algorithm for precision medicine – Matt Might

Matt Might, Harvard Medical School | Wednesday, September 30, 2015 | Video


President Obama recently launched the Precision Medicine Initiative, a confluence of efforts in data science, bioinformatics, systems biology and genomics. Precision medicine’s promise of “the right medicine to the right patient at the right time” is predicated on the assumption that a patient’s health data may be mapped directly to the “right medicine.” It is reasonable to assume that such a mapping exists (in theory), but it is not yet clear how complex the implementation of that mapping will become. With the claim that genomic data will be a key driver in precision medicine, rare genetic disorders offer a window into the genome-guided aspects of precision medicine. This talk provides a cautionary yet optimistic portrait of what full-scale precision medicine will entail, illustrated by the speaker’s first-hand experience with aftermath of the discovery that his son was the first known patient of a novel and ultra-rare genetic disorder — NGLY1 deficiency.


Matt Might is an associate professor of computer science at the University of Utah and a visiting associate professor in computer science at the Harvard Medical School.

Local views and global conclusions – Nati Linial

Nati Linial, Hebrew University of Jerusalem | Thursday, September 17, 2015 | Video


We start by describing a challenge in bioinformatics to illustrate a very universal phenomenon: The Protein Interaction graph G of an organism has one vertex for each of its proteins and an edge for each pair of interacting proteins. Several competing theories attempt to describe how such graphs emerge in evolution and we wish to tell which theory provides a better explanation.

A major difficulty in resolving such problems is that G is huge so it is unrealistic to calculate most of its nontrivial graph parameters. But even a huge graph G can be efficiently sampled. Given a small integer k (say k=10), the k-profile of G is a distribution on k-vertex graphs. It is derived by randomly sampling k vertices in G and observing the subgraph that they induce. A theory largely developed in MSR (“Theory of graph limits” – Lovasz, Szegedy, Chayes, Borgs, Cohn, Friedman…) offers a clue. It says essentially that to decide whether a series of large graphs is derived from a given statistical model it is enough to check that the graphs’ profiles behave as they should.

I will give you some sense of the theory of graph limits and then move to discuss profiles. The two main questions are: (i) Which profiles are possible? (ii) What global properties of G can you derive, based on its profiles?

Personalized Health with Gaussian Processes – Neil Lawrence

Neil Lawrence, University of Sheffield | Wednesday, August 19, 2015 | Video


Modern data connectivity gives us different views of the patient which need to be unified for truly personalized health care. I’ll give an personal perspective on the type of methodological and social challenges we expect to arise in this this domain and motivate Gaussian process models as one approach to dealing with the explosion of data.


Neil Lawrence received his bachelor’s degree in Mechanical Engineering from the University of Southampton in 1994. Following a period as an field engineer on oil rigs in the North Sea he returned to academia to complete his PhD in 2000 at the Computer Lab in Cambridge University. He spent a year at Microsoft Research in Cambridge before leaving to take up a Lectureship at the University of Sheffield, where he was subsequently appointed Senior Lecturer in 2005. In January 2007 he took up a post as a Senior Research Fellow at the School of Computer Science in the University of Manchester where he worked in the Machine Learning and Optimisation research group. In August 2010 he returned to Sheffield to take up a collaborative Chair in Neuroscience and Computer Science.

Neil’s main research interest is machine learning through probabilistic models. He focuses on both the algorithmic side of these models and their application. He has a particular focus on applications in personalized health and computational biology, but happily dabbles in other areas such as speech, vision and graphics.

Neil was Associate Editor in Chief for IEEE Transactions on Pattern Analysis and Machine Intelligence (from 2011-2013) and is an Action Editor for the Journal of Machine Learning Research. He was the founding editor of the JMLR Workshop and Conference Proceedings (2006) and is currently series editor. He was an area chair for the NIPS conference in 2005, 2006, 2012 and 2013, Workshops Chair in 2010 and Tutorials Chair in 2013. He was General Chair of AISTATS in 2010 and AISTATS Programme Chair in 2012. He was Program Chair of NIPS in 2014 and is General Chair for 2015.

Can We Agree on Science? Measuring the Ideological Alignment of Science with Book Co-purchase Data – James Evans

James Evans, University of Chicago | Wednesday, August 12, 2015


Does science constitute a “public sphere” for reasoned debate in the United States? Attacks on science in the media and the liberal credentials of most scientists could suggest no, but recent surveys find the public does not regard science as liberal and overwhelmingly acknowledges scientific contributions to society. We used millions of recommendations based on co-purchases between political and scientific books as a behavioral indicator for whether science bridges or deepens political divides. Findings reveal that books from the social sciences and hot-button fields (e.g., climatology) are most politically relevant, but books from general scientific disciplines (e.g., physics, astronomy, and zoology) are more co-purchased with liberal books, while those in practical, commercially relevant fields (e.g., medicine, criminology, and geology) are more co-purchased with conservative books. Moreover, liberal books tend to be co-purchased with a much broader sample of science books, indicating that conservatives have more selective interest in science. We conclude that the political left and right share an interest in science in general, but not science in particular.


James Evans is Director of Knowledge Lab (, senior fellow at the Computation Institute, associate professor of Sociology and the College, and member of the Committee on Conceptual and Historical Studies of Science at the University of Chicago. He is founding director of the Masters program in Computational Social Science (starting 2016) at the University of Chicago. His research focuses on the collective system of thinking and knowing, ranging from the distribution of attention and intuition, the origin of ideas and shared habits of reasoning to processes of agreement (and dispute), accumulation of certainty (and doubt), and the texture–novelty, ambiguity, topology–of human understanding. Evans is especially interested in innovation–how new ideas and practices emerge–and the role that social and technical institutions (e.g., the Internet, markets, collaborations) play in collective cognition and discovery. Much of Evans work has focused on areas of modern science and technology, but he is also interested in other domains of knowledge–news, law, religion, gossip, hunches and historical modes of thinking and knowing. Evans supports the creation of novel observatories for human understanding and action through crowd sourcing, information extraction from text and images, and the use of distributed sensors (e.g., RFID tags, cell phones). He uses machine learning, generative modeling, social and semantic network representations to explore knowledge processes, scale up interpretive and field-methods, and create alternatives to current discovery regimes. His research is funded by the National Science Foundation, the National Institutes of Health, the Templeton Foundation and other sources, and has been published in Science, American Journal of Sociology, American Sociological Review, Social Studies of Science, Administrative Science Quarterly, PLoS Computational Biology and other journals. My work has been featured in Nature, the Economist, Atlantic Monthly, Wired, NPR, BBC, El País, CNN and many other outlets.

Learning and Efficiency in Games with Dynamic Population – Eva Tardos

Eva Tardos, Cornell | Wednesday, July 29, 2015 | Video


Selfish behavior can often lead to suboptimal outcome for all participants. This is especially true in dynamically changing environments where the game or the set of the participants can change at any time without even the players realizing it. Over the last decade we have developed good understanding how to quantify the impact of strategic user behavior on overall performance via studying via equilibria of the games. In this talk we will consider the quality of outcomes in games when the population of players is dynamically changing, and where participants have to adapt to the dynamic environment. We show that in large classes of games (including congestion games), if players use a form of learning that helps them to adapt to the changing environment, this guarantees high social welfare, even under very frequent changes. A main technical tool for our analysis is a connection between differential privacy and high efficiency of learning outcomes in frequently changing repeated games. Joint work with Thodoris Lykouris and Vasilis Syrgkanis.


Éva Tardos is a Jacob Gould Schurman Professor of Computer Science Professor, at Cornell University, and was department chair 2006-2010. She received her BA and PhD from Eotvos University in Budapest. She has been elected to the National Academy of Engineering, the National Academy of Sciences, the American Academy of Arts and Sciences, is an external member of the Hungarian Academy of Sciences, and is the recipient of a number of fellowships and awards including the Packard Fellowship, the Goedel Prize, Dantzig Prize, Fulkerson Prize, and the IEEE Technical Achievement Award. She was editor editor-in-Chief of SIAM Journal of Computing 2004-2009, and is currently editor of several other journals including the Journal of the ACM and Combinatorica, served as problem committee member and chair for many conferences.

Tardos’s research interest is algorithms and algorithmic game theory, the subarea of theoretical computer science theory of designing systems and algorithms for selfish users. Her research focuses on algorithms and games on networks. She is most known for her work on network-flow algorithms, approximation algorithms, and quantifying the efficiency of selfish routing.

Privacy Protection, Personalized Medicine and Genetic Testing – Catherine Tucker

Catherine Tucker, MIT | Wednesday, July 15, 2015 | Video


Personalized medicine – where the treatment is as individual as the patient – has been discussed as the future of medicine. But the use of a person’s genetic code to personalize treatment raises new and difficult privacy concerns. Professor Tucker will discuss current approaches to regulating genetic privacy at the state level and the degree of success such approaches have had at promoting the spread of personalized medicine and testing.


Catherine Tucker is the Mark Hyman Jr. Career Development Professor and Associate Professor of Management Science at MIT Sloan. Her research interests lie in how technology allows firms to use digital data to improve their operations and marketing and in the challenges this poses for regulations designed to promote innovation. She has particular expertise in online advertising, digital health, social media, and electronic privacy. Generally, most of her research lies in the interface between Marketing, Economics and Law. She has received an NSF CAREER award for her work on digital privacy, the Erin Anderson Award for Emerging Marketing Scholar and Mentor, the Paul E. Green Award for contributions to the practice of Marketing Research and a Garfield Award for her work on electronic medical records. She has testified before Congress on privacy regulation, as well as presenting her research on privacy to the FCC, FTC and OECD. In addition to her work on privacy and digital data, she has also written extensively on how the online and technology environment changes and challenges intellectual property regimes in the sphere of patent assertion entities, trademarks used as search terms, and copyright issues for online aggregators. Her more practitioner-oriented research in marketing tackles the challenge of how to design online advertising campaigns which do not appear intrusive to the viewer, and have the potential to be spread virally.

Dr. Tucker is Associate Editor at Management Science, Co-Editor at Quantitative Marketing and Economics and Co-Editor of the recent NBER volume on the Economics of Digitization. She is a Research Associate at the National Bureau of Economic Research.

She teaches MIT Sloan’s MBA Elective on `Pricing’ and the Executive MBA course `Marketing Management for the Senior Executive’. She also teaches in various specialized executive education programs on entrepreneurship, creating thriving platforms ecosystems and innovation. She has received the Jamieson Prize for Excellence in Teaching as well as being voted `Teacher of the Year’ at MIT Sloan.

She holds a PhD in economics from Stanford University, and a BA in Politics, Philosophy and Economics from Oxford University.

Unbalanced Random Matching Markets: The Stark Effect of Competition – Itai Ashlagi

Itai Ashlagi, MIT | Wednesday, July 1, 2015 | Video


Stability is used often as a criterion in organizing clearinghouses for two-sided matching markets, where agents on both sides of the market have preferences over potential matches. We study competition in matching markets with random heterogeneous preferences by considering markets with an unequal number of agents on either side. First, we show that even the slightest imbalance yields an essentially unique stable matching. Second, we give a tight description of stable outcomes, showing that matching markets are extremely competitive. Each agent on the short side of the market is matched to one of his top preferences and each agent on the long side does almost no better than being matched to a random partner. Our results suggest that any matching market is likely to have a small core, explaining why empirically small cores are ubiquitous and solving a longstanding puzzle.


Itai Ashlagi is an Assistant Professor of Operations Management at Sloan, MIT. He graduated from the Technion and did his postdoc at Harvard Business School. Ashlagi is mainly interested in market design. He received the outstanding paper award in the ACM conference of Electronic Commerce and the NSF Career award. He is a Franz Edelman laureate for his work on kidney exchange, which shaped policies of numerous kidney exchange programs. He will join MS&E at Stanford in the fall of 2015.

Some limitations and possibilities toward data-driven optimization – Yaron Singer

Yaron Singer, Harvard | Wednesday, June 24, 2015 | Video


As we grow highly dependent on big data for making predictions, we translate these predictions into models that help us make informed decisions. But how do the guarantees we have on predictions translate to guarantees on decisions? In many cases, we learn models from sampled data and then aim to use these models to make decisions. In some cases, despite having access to large data sets, the current frameworks we have for learnability do not suffice to guarantee desirable outcomes. In other cases, the learning techniques we have introduce estimation errors which can result in poor outcomes and stark impossibility results. In this talk we will formalize some of these ideas using convex and combinatorial optimization.


Yaron Singer is an Assistant Professor of Computer Science at Harvard University. He was previously a postdoctoral researcher at Google Research and obtained his PhD from UC Berkeley. He is the recipient of the NSF CAREER award, 2012 Best Student Paper Award at the ACM conference on Web Search and Data Mining, the 2010 Facebook Fellowship, the 2009 Microsoft Research Fellowship, and several awards for entrepreneurial work on social networks.

Learning, mixing, and complexity- a free ride on the second law – James Lee

James Lee, University of Washington | Monday, June 1, 2015 | Video


The principle of maximum entropy states that, given some data, among all hypothetical probability distributions that agree with the data, the one of maximum entropy best represents the current state of knowledge. It might be natural to expect that this philosophy often yields a “simple” hypothesis since it tries to avoid making the hypothesis more informative than it deserves to be.

Viewing entropy as an additional resource to be optimized is an extremely powerful idea with a wide range of applications (and a correspondingly large array of names: boosting, entropy-regularized gradient descent, multiplicative weights update, log-Sobolev inequalities, Gibbs measures, etc.).

I will focus specifically on the role of entropy maximization in encouraging simplicity. This has a number of surprising applications in discrete mathematics and the theory of computation. We’ll see three instantiations of this principle: in additive number theory, functional analysis, and complexity theory. For the last application, it will turn out that one needs to extend max-entropy to the setting of quantum information and von Neumann entropy.

The philosophy and applications will be discussed at a high level suitable for a general scientific audience.


James R. Lee is an Associate Professor of Computer Science at the University of Washington. His research leverages tools from probability and analysis to attack fundamental problems in discrete mathematics, algorithms, and complexity theory. His work has been recognized by an NSF CAREER award, a Sloan Research Fellowship, and recently a best paper award at STOC 2015 (with P. Raghavendra and D. Steurer) for an application of entropy maximization to computational lower bounds.

Better Science Through Better Bayesian Computation – Ryan Adams

Ryan Adams, Harvard | Wednesday, May 20, 2015


As we grapple with the hype of “big data” in computer science, it is important to remember that the data are not the central objects: we collect data to answer questions and inform decisions in science, engineering, policy, and beyond. In this talk, I will discuss my work in developing tools for large-scale data analysis, and the scientific collaborations in neuroscience, chemistry, and astronomy that motivate me and keep this work grounded. I will focus on two lines of research that I believe capture an important dichotomy in my work and in modern probabilistic modeling more generally: identifying the “best” hypothesis versus incorporating hypothesis uncertainty. In the first case, I will discuss my recent work in Bayesian optimization, which has become the state-of-the-art technique for automatically tuning machine learning algorithms, finding use across academia and industry.

In the second case, I will discuss scalable Markov chain Monte Carlo and the new technique of Firefly Monte Carlo, which is the first provably correct MCMC algorithm that can take advantage of subsets of data.


Ryan Adams is an Assistant Professor of Computer Science at Harvard. He received his Ph.D. in Physics at Cambridge as a Gates Scholar. He was a CIFAR Junior Research Fellow at the University of Toronto before joining the faculty at Harvard. He has won paper awards at ICML, AISTATS, and UAI, and his Ph.D. thesis received Honorable Mention for the Savage Award for Theory and Methods from the International Society for Bayesian Analysis. He also received the DARPA Young Faculty Award and the Sloan Fellowship. Dr. Adams is the CEO of Whetlab, a machine learning startup, and co-hosts the popular Talking Machines podcast.

Assessing the Creepy Factor: Shifting from regulatory ethics models to more proactive approaches to ‘doing the right thing’ in technology research – Annette Markham

Annette Markham, Aarhus University | Wednesday, May 6, 2015 | Video


What constitutes ethical design of technologies, ethical use of data, and ethical research? How can we pay better attention to the ways in which some aspects of our research or outcomes of our designs might seem ‘creepy’?

In this talk, I begin with the premise that “doing the right thing” is an outcome of rhetorically powerful tangles of human and non-human elements, embedded in deep—often invisible—structures of software, politics, and habits. Every action by individuals—whether designers, programmers, marketers, researchers, policy makers or consumers—reinforces, resists, and reconfigures existing ethical boundaries for what is acceptable and just.

Despite the development of nuanced approaches for ethics in digital and technology studies, the general language surrounding ethics has remained ensconced in that of regulations, requirements, and concepts, born from biomedical models that don’t fit well with contemporary research environments and practices. In this talk, I suggest a framework of ethics in digital research that focuses less on ‘ethics’ and more on what might be potentially ‘creepy’ about what we’re doing in our everyday research and design. This is combined with a future oriented ‘what if’ approach. Placing more responsibility on one’s personal choices is not the most comfortable position, but as the world grows more technologically mediated and digitally saturated, it is particularly important to speculate about future possibilities and harms.

I hope to conclude this talk by introducing and getting feedback on sample scenarios that could be used to help Microsoft Researchers include ethical considerations in both conceptual and practical research contexts.


Annette Markham is Associate Professor of Information Studies at Aarhus University in Denmark and Affiliate Professor of Digital Ethics in the School of Communication at Loyola University in Chicago. She earned her PhD in organizational communication (Purdue University, 1998), with a strong emphasis in interpretive, qualitative, and ethnographic methods. Annette’s early research focused on how identity, relationships, and cultural formations constructed in and influenced by digitally saturated socio-technical contexts. Her pioneering sociological work related to digital identity is well represented in her book Life Online: Researching real experience in virtual space (Altamira 1998). Her more recent research focuses on innovative qualitative methodologies for studying networked sociality and ethics of social research and interaction design. Her work can be found in a range of international journals, handbooks, and edited collections, including the book Internet Inquiry (2009, co-edited with Nancy Baym).

The Limits of Reputation in Platform Markets: An Empirical Analysis and Field Experiment – Steve Tadelis

Steve Tadelis, Berkeley | Wednesday, April 22, 2015


Reputation mechanisms used by platform markets suffer from two problems. First, buyers may draw conclusions about the quality of the platform from single transactions, causing a reputational externality. Second, reputation measures may be coarse or biased, preventing buyers from making proper inferences. We document these problems using eBay data and claim that platforms can benefit from identifying and promoting higher quality sellers. Using an unobservable measure of seller quality we demonstrate the benefits of our approach through a large-scale controlled experiment. Highlighting the importance of reputational externalities, we chart an agenda that aims to create more realistic models of platform markets.


These days my research primarily revolves around e-commerce and the economics of the internet. During the 2011-2013 academic years I was on leave at eBay research labs, where I hired and led a team of research economists. Our work focused on the economics of e-commerce, with particular attention to creating better matches of buyers and sellers, reducing market frictions by increasing trust and safety in eBay’s marketplace, understanding the underlying value of different advertising and marketing strategies, and exploring the market benefits of different pricing structures. Aside from the economics of e-commerce, my main fields of interest are the economics of incentives and organizations, industrial organization, and microeconomics. Some of my past research aspired to advance our understanding of the roles played by two central institutions—firms and contractual agreements—and how these institutions facilitate the creation of value. Within this broader framework, I explored firm reputation as a valuable, tradable asset; the effects of contract design and organizational form on firm behavior with applications to outsourcing and privatization; public and private sector procurement and award mechanisms; and the determinants of trust.

Measuring Rhetoric: Statistical Language Models in Social Science – Matt Taddy

Matt Taddy, University of Chicago | Wednesday, April 8, 2015 | Video


Social scientists are embracing the idea of using `text as data’ as a way to quantify and evaluate social theories. I’ll discuss a brief history of how this strategy has worked and evolved, and pitch some new approaches for combining social measurement with state-of-the-art natural language processing. We’ll focus on the massive multinomial regression models that serve as a basis for text analysis and the distributed computing strategies that allow inference on truly Big Data. I’ll then work through a number of examples of social science questions being asked and answered via statistical NLP, with data from online reviews on Yelp, the US congressional record, and communications between buyers and sellers on eBay.


Matt Taddy is Associate Professor of Econometrics and Statistics at the University of Chicago Booth School of Business. His research is focused on statistical methodology and data mining, driven by applications in business and engineering. He developed and teaches the MBA ‘Big Data’ course at Chicago Booth.

Taddy works on building robust solutions for large scale data analysis problems, at the interface of econometrics and machine learning. This involves dimension reduction techniques for massive datasets and development of models for inference on the output of these algorithms. He has collaborated both with small start-ups and with large research agencies, including NASA Ames, and Lawrence Livermore, Sandia, and Los Alamos National Laboratories, and is a scientist at eBay research labs.

Taddy earned his PhD in Applied Math and Statistics in 2008 from the University of California, Santa Cruz, as well as a BA in Philosophy and Mathematics and an MSc in Mathematical Statistics from McGill University. He joined the Chicago Booth faculty in 2008.

The Eureka Myth: Creators, Innovators and Everyday Intellectual Property – Jessica Silbey

Jessica Silbey, Suffolk | Wednesday, March 11, 2015 | Video


Are innovation and creativity helped or hindered by our intellectual property laws? In the two hundred plus years since the Constitution enshrined protections for those who create and innovate, we’re still debating the merits of IP laws and whether or not they actually work as intended. Artists, scientists, businesses, and the lawyers who serve them, as well as the Americans who benefit from their creations all still wonder: what facilitates innovation and creativity in our digital age? And what role, if any, do our intellectual property laws play in the growth of innovation and creativity in the United States?

Incentivizing the “progress of science and the useful arts” has been the goal of intellectual property law since our constitutional beginnings. The Eureka Myth cuts through the current debates and goes straight to the source: the artists and innovators themselves. Silbey makes sense of the intersections between intellectual property law and creative and innovative activity by centering on the stories told by artists, scientists, their employers, lawyers and managers, describing how and why they create and innovate and whether or how IP law plays a role in their activities. Their employers, business partners, managers, and lawyers also describe their role in facilitating the creative and innovative work. Silbey’s connections and distinctions made between the stories and statutes serve to inform present and future innovative and creative communities.

Breaking new ground in its examination of the U.S. economy and cultural identity, The Eureka Myth draws out new and surprising conclusions about the sometimes misinterpreted relationships between creativity and intellectual property protections.


Professor Jessica Silbey teaches at Suffolk University Law School in Boston in the areas of intellectual property and constitutional law. Professor Silbey received her B.A. from Stanford University and her J.D. and Ph.D. (Comparative Literature) from the University of Michigan. After clerking for Judge Robert E. Keeton on the United States District Court for the District of Massachusetts and Judge Levin Campbell on the United States Court of Appeals for the First Circuit, she practiced law in the disputes department of the Boston office of Foley Hoag LLP focusing on intellectual property, bankruptcy and reproductive rights. Professor Silbey’s scholarly expertise is in the cultural analysis of law, exploring the law beyond its doctrine to the contexts and processes in which legal relations develop and become significant for everyday actors. In the field of intellectual property, Professor Silbey’s scholarship focuses on the humanistic and sociological dimensions of the legal regulation of creative and innovative work. Some of her IP publications include The Eureka Myth: Creators, Innovators and Everyday Intellectual Property (Stanford University Press 2014); Patent Variation: Discerning Diversity Among Patent Functions, 45 Loy. U. Chi. L. Rev. 441 (2013); Harvesting Intellectual Property, ‘Inspired Beginnings and ‘Work Makes Work’: Two Stages in the Creative Process of Artists and Innovators, 86 Notre Dame L. R. 2091 (2011), Comparative Tales of Origins and Access: The Future of Intellectual Property Law, 61 Case Wes. Res. L. R. 195 (2011), and Mythical Beginnings of Intellectual Property, 15 Geo. Mason L. R. 319 (2008). Professor Silbey has also published widely in the field of law and film, exploring how film is used as a legal tool and how it becomes an object of legal analysis in light of its history as a cultural object and art form. Representative publications include Law and Justice on the Small Screen (Hart, 2012) (with Peter Robson); Evidence Verité and the Law of Film, 31 Cardozo L. R. 1257 (2010); Cross-Examining Film, 8 U. Md. J. Race, Religion & Gender & L. 101 (2009); and Judges as Film Critics: New Approaches to Filmic Evidence, 39 Mich. J. L. Reform 493 (2004).

After Math: Following Mathematics into the Digital – Stephanie Dick

Stephanie Dick, Harvard | Wednesday, March 4, 2015 | Video


The advent of modern digital computing in the mid-twentieth century precipitated many transformations in the practices of mathematical knowledge production. However, early computing practitioners throughout the United States subscribed to complicated and conflicting visions of just how much the computer could contribute to mathematics – each suggesting a different division of mathematical labor between humans and computers and a hierarchization of the tasks involved. Some imagined computers as mere plodding “slaves” who would take over tedious and mechanical elements of mathematical research. Others imagined them more generously as “mentors” or “collaborators” that could offer novel insight and direction to human mathematicians. Still others believed that computers would eventually become autonomous agents of mathematical research. And computing communities did not simply imagine the potential of the computer differently; they also built those different visions right in to computer programs that enabled new ways of doing mathematics with computers. With a focus on communities based in the United States in the second half of the twentieth century, this talk will explore different visions of the computer as a mathematical agent, the software that was crafted to animate those imaginings, and the communities and practices of mathematical knowledge-making that emerged in tandem.


Stephanie Dick is a Junior Fellow with the Harvard Society of Fellows. She recently completed a PhD in the Department of History of Science at Harvard University. Her work explores the history of mathematics and computing in the postwar United States. She focuses on the history of mathematical software and its epistemological significance.

Perspectives on Recombination – Elizabeth Pontikes

Elizabeth Pontikes, University of Chicago | Wednesday, February 25, 2015


Research in economics and sociology over the past century has pointed to recombination as the source for novel social and economic developments. This study suggests that the categorical structure a person uses to understand a domain is fundamental to this concept. This is studied in an investigation of venture capital financing of software organizations. Findings show that venture capitalists are more likely to invest in companies that engage in recombination based on market categories, but that traditional measures of recombination based on patent classes do not have predictive value. Results are strongest for private equity venture capitalists and weakest for corporate venture capitalists, suggesting that people who value novelty based on breaking down existing boundaries will favor recombination, while those who prefer progress that reinforces existing categories will avoid it.


Elizabeth Pontikes is an Associate Professor of Organizations and Strategy at the University of Chicago Booth School of Business. Her research focuses on market classification, innovation, and knowledge development. In her research, Pontikes shows that in systems of market classification, categories vary in how constraining they are, and that category leniency affects how organizational members are evaluated, and when new categories emerge in a classification system. She has studied these ideas in the context of the software industry, the computer industry, and also has studied negative categorization in the context of the Red Scare in Hollywood. In addition, Pontikes is currently working on a project that investigates how rap artists define their identity in their lyrics, and how this is received by both popular and critical audiences. Pontikes has been published in a number of scholarly journals including Administrative Science Quarterly, American Sociological Review, Management Science, and Sociological Science.

Optimal Design for Social Learning – Johannes Horner

Johannes Horner, Yale | Wednesday, December 3, 2014 | Video


We study the design of a recommender system for organizing social learning on a product. The optimal design trades off fully transparent social learning to improve incentives for early experimentation, by selectively over-recommending a product in the early phase of the product release. Under the optimal scheme, experimentation occurs faster than under full transparency but slower than under the first-best opti- mum, and the rate of experimentation increases over an initial phase and lasts until the posterior becomes sufficiently bad in which case the recommendation stops along with experimentation on the product. Fully transparent recommendation may become optimal if the (socially-benevolent) designer does not observe the agents’ costs or the agents choose the timing of receiving a recommendation.


Johannes Hörner is Professor of Economics, Department of Economics, and Cowles Foundation for Research in Economics, Yale University. He has received his Ph.D. in economics from the University of Pennsylvania in 2000, and has held previous positions at the Kellogg School of Management, Northwestern University (2000–2008).His academic interests range from game theory to the theory of industrial organization. His research has focused on repeated games, dynamic games, and auctions.

Physics-inspired algorithms and phase transitions in community detection – Christopher Moore

Cristopher Moore, Santa Fe Institute | Tuesday, November 18, 2014 | Video


Detecting communities, and labeling nodes, is a ubiquitous problem in the study of networks. Recently, we developed scalable Belief Propagation algorithms that update probability distributions of node labels until they reach a fixed point. In addition to being of practical use, these algorithms can be studied analytically, revealing phase transitions in the ability of any algorithm to solve this problem. Specifically, there is a detectability transition in the stochastic block model, below which no algorithm can label nodes better than chance. This transition was subsequently established rigorously by Mossel, Neeman, and Sly, and Massoulie.

I’ll explain this transition, and give an accessible introduction to Belief Propagation and the analogy with free energy and the cavity method of statistical physics. We’ll see that the consensus of many good solutions is a better labeling than the “best” solution — something that is true for many real-world optimization problems. While many algorithms overfit, and find “communities” even in random graphs where none exist, our method lets us focus on statistically-significant communities. In physical terms, we focus on the free energy rather than the ground state energy.I’ll then turn to spectral methods. It’s popular to classify nodes according to the first few eigenvectors of the adjacency matrix or the graph Laplacian. However, in the sparse case these operators get confused by localized eigenvectors, focusing on high-degree nodes or dangling trees rather than large-scale communities. As a result, they fail significantly above the detectability transition. I will describe a new spectral algorithm based on the non-backtracking matrix, which avoids these localized eigenvectors: it appears to be optimal in the sense that it succeeds all the way down to the transition. Making this rigorous will require us to prove an interesting conjecture in the theory of random matrices and random graphs.

This is joint work with Aurelien Decelle, Florent Krzakala, Elchanan Mossel, Joe Neeman, Mark Newman, Allan Sly, Lenka Zdeborova, and Pan Zhang.


Cristopher Moore is a Professor at the Santa Fe Institute. He received his B.A. in Physics, Mathematics, and Integrated Science from Northwestern University, and his Ph.D. in Physics from Cornell. In 2000, he joined the University of New Mexico faculty, with joint appointments in Computer Science, and Physics and Astronomy. In 2012, Moore left the University of New Mexico and became full-time resident faculty at the Santa Fe Institute. He has published over 120 papers at the boundary between physics and computer science, ranging from quantum computing, to phase transitions in NP-complete problems, to the theory of social networks and efficient algorithms for analyzing their structure. With Stephan Mertens, he is the author of The Nature of Computation, published by Oxford University Press.

Mapping single cells: A geometric approach – Dana Pe'er

Dana Pe’er, Columbia | Wednesday, November 5, 2014 | Video


High dimensional single cell technologies are on the rise, rapidly increasing in accuracy and throughput. These offer computational biology both a challenge and an opportunity. One of the big challenges with this data-type is to understand regions of density in this multi-dimensional space, given millions of noisy measurements. Underlying many of our approaches is mapping this high-dimensional geometry into a nearest neighbor graph and characterization single cell behavior using this graph structure. We will discuss a number of approaches (1) An algorithm that harnesses the nearest neighbor graph to order cells according to their developmental maturity and its use to identify novel progenitor B-cell sub-populations. (2) Using reweighted density estimation to characterize cellular signal processing in T-cell activation. (2) New clustering and dimensionality reduction approaches to map heterogeneity between cells; with an application to characterizing tumor heterogeneity in Acute Myeloid Leukemia.


Dana Pe’er is an associate professor in the Departments of Biological Sciences and Computer Science. Her lab endeavors to understand the organization, function, and evolution of molecular networks, particularly how variation in DNA sequence alters regulatory networks and leads to the vivid phenotypic diversity of life. Her team develops computational methods that integrate diverse high-throughput data to provide a holistic, systems-level view of molecular networks. She is particularly interested in exploring how systems biology can be used to personalize care for people with cancer. By developing models that can predict how individual tumors will respond to certain drugs and drug combinations, her goal is to develop ways to determine the best drug regime for each patient. Her interest is not only in understanding which molecular components go wrong in cancer cells, but also in using this information to improve cancer therapeutics.

Dr. Pe’er is the recipient of the 2014 Overton Prize, and has been recognized with the Burroughs Wellcome Fund Career Award, an NIH Directors New Innovator Award, an NSF CAREER Award, and a Stand Up To Cancer Innovative Research Grant. She was also named a Packard Fellow in Science and Engineering.

Barack Obama and the politics of social media for national policy-making – James Katz

James Katz, Boston University | Wednesday, October 15, 2014


Social media help people do most everything, ranging from meeting new friends and finding new restaurants to overthrowing dictatorships. This includes political campaigning; one need look no further than Barack Obama’s successful presidential campaigns to see how these communication technologies can alter the way politics is conducted. Yet social media have not had much import for setting national policy as part of regular administrative routines. This is the case despite the fact that, since his election in 2008, President Obama has on several occasions proclaimed that he wanted his administration to draw on social media to make the federal government run better. While there have been some modifications to governmental procedures due to the introduction of social media, the Obama administration practices have fallen far short of its leader’s audacious vision. Despite voluminous attention to social media in other spheres of activity, there has been little to point to in terms of successfully drawing on the public to help set national policies. What might account for this? I try to answer this question in my talk by exploring the attempts by the Obama White House to use social media tools and the consequences arising from such attempts. I also suggest some potential reasons behind the particular uses and outcomes that have emerged in terms of presidential-level social media outreach. As part of my conclusion, I outline possible future directions.


James E. Katz, Ph.D., is the Feld Family Professor of Emerging Media at Boston University’s College of Communication where he directs its Center for Mobile Communication Studies and Division of Emerging Media. His research on the internet, social media and mobile communication has been internationally recognized, and he is frequently invited to address high-level industry, governmental and academic groups on his research findings. His latest book, with Barris and Jain, is The Social Media President: Barack Obama and the Politics of Citizen Engagement on which this talk is based.

Cooperation on Social Networks – Nageeb Ali

Nageeb Ali, UCSD | Wednesday, October 1, 2014 | Video


At most places, and at most times, cooperation takes place in the absence of legal or contractual enforcement. What motivates players to cooperate? A growing literature in the social sciences emphasizes the importance of future interactions and social mechanisms by which defectors are punished both by their victims and third-parties. This perspective has, in recent years, influenced our understanding of contractual and lending relationships in developing economies, reputations in market platforms such as eBay, and even that of indirect reciprocity in theoretical biology. In this talk, I will describe how the nature and strength of these incentives varies with a social network, how a player may cooperate so as to preserve his reputation in a social network, and what guarantees that a victim of defection truthfully reveals to others that someone else has violated the social norm. We will see that dividing society into cliques and that a modicum of forgiveness can facilitate cooperation. We might see that a commonly made assumption made in much of the literature on cooperation—that victims always reveal when someone else has defected—may be less innocuous than it seems.


S. Nageeb Ali is an assistant professor of economics at UCSD. He studies game-theoretic models of cooperation, social learning, political economy, and behavioral economics. He received his Ph.D. from Stanford University in 2007, and is a frequent Microsoft visitor.

The origins of common sense: Modeling human intelligence with probabilistic programs and program induction – Joshua Tenenbaum

Joshua Tenenbaum, MIT | Wednesday, September 17, 2014 | Video


Our work seeks to understand the roots of human thinking by looking at the core cognitive capacities and learning mechanisms of young children and infants. We build computational models of these capacities with the twin goals of explaining human thought in more principled, rigorous “reverse engineering” terms, and engineering more human-like AI and machine learning systems. This talk will focus on two ways in which the intelligence of very young children goes beyond the conventional paradigms in machine learning: (1) Scene understanding, where we cannot detect not only objects and their locations, but what is happening, what will happen next, who is doing what to whom and why, in terms of our intuitive theories of physics (forces, masses) and psychology (beliefs, desires, …); (2) Learning concepts from examples, where just a single example is often sufficient to grasp a new concept and generalize in richer ways than machine learning systems can typically do even with hundreds or thousands of examples. I will show how we are beginning to capture these reasoning and learning abilities in computational terms using techniques based on probabilistic programs and program induction, embedded in a broadly Bayesian framework for inference under uncertainty.


Josh Tenenbaum studies learning, reasoning and perception in humans and machines, with the twin goals of understanding human intelligence in computational terms and bringing computers closer to human capacities. His current work focuses on building probabilistic models to explain how people come to be able to learn new concepts from very sparse data, how we learn to learn, and the nature and origins of people’s intuitive theories about the physical and social worlds. He is Professor of Computational Cognitive Science in the Department of Brain and Cognitive Sciences at MIT, and is a member of the Computer Science and Artificial Intelligence Laboratory (CSAIL). He received his Ph.D. from MIT in 1999, and was a member of the Stanford University faculty in Psychology and (by courtesy) Computer Science from 1999 to 2002. His papers have received awards at numerous conferences, including CVPR (the IEEE Computer Vision and Pattern Recognition conference), ICDL (the International Conference on Learning and Development), NIPS, UAI, IJCAI and the Annual Conference of the Cognitive Science Society. He is the recipient of early career awards from the Society for Mathematical Psychology (2005), the Society of Experimental Psychologists, and the American Psychological Association (2008), and the Troland Research Award from the National Academy of Sciences (2011).

Robust Probabilistic Inference – Yishay Mansour

Yishay Mansour, MSR Israel | Wednesday, August 27, 2014 | Video


Probabilistic Inference is the task of given a certain set of observations, to deduce the probability of various outcomes. This is a very basic task both in statistics and in machine learning.Robust probabilistic inference is an extension of probabilistic inference, where some of the observations are adversarially corrupted. Examples of where such a model may be relevant are spam detection, where spammers try adversarially to fool the spam detectors, or failure detection and correction, where the failure can be modeled as a “worse case” failure. The framework can be also used to model selection between a few alternative models that possibly generate the data.Technically, we model robust probabilistic inference as a zero-sum game between an adversary, who can select a modification rule, and a predictor, who wants to accurately predict the state of nature. Our main result is an efficient near optimal algorithm for the robust probabilistic inference problem. More specifically, given a black-box access to a Bayesian inference in the classic (adversary-free) setting, our near optimal policy runs in polynomial time in the number of observations and the number of possible modification rules.

This is a joint work with Aviad Rubinstein and Moshe Tennenholtz.


Prof. Yishay Mansour got his PhD from MIT in 1990, following it he was a postdoctoral fellow in Harvard and a Research Staff Member in IBM T. J. Watson Research Center. Since 1992 he is at Tel-Aviv University, where he is currently a Professor of Computer Science and has served as the head of the School of Computer Science during 2000-2002. Prof. Mansour has held visiting positions with Bell Labs, AT&T research Labs, IBM Research, and Google Research. Prof. Mansour has published over 50 journal papers and over 100 proceeding paper in various areas of computer science with special emphasis on communication networks machine learning, and algorithmic game theory. Prof. Mansour is currently an associate editor in a number of distinguished journals and has been on numerous conference program committees. He was both the program chair of COLT (1998) and served on the COLT steering committee. He has supervised over a dozen graduate students in various areas including communication networks, machine learning, algorithmic game theory and theory of computing.

Nerves and Synapses - A General Preview – Michal Linial

Michal Linial, Hebrew University | Wednesday, August 13, 2014 | Video


My talk is a brief preview of neuroscience (pre-101..). I will share with you some of the brain’s mysteries and will illustrate the capacity of neurons to rewire and thus to learn (and forget). To do so, we will discuss (briefly) how neurons convey information, what are the principles underlying neuronal communication and the fundamental rules of electrical and chemical messengers. The uniformity and the variability of neurons that are involved in high brain functions (mathematics?) and those that make sure that we quickly remove our finger from a hot plate will be discussed. I will mention the capacity of the human brain vis-à-vis that of our cousins, the chimps, and other nerve systems. Is our brain really so different? (Probably so), what makes us human? (I have no clue..), why are we all fascinated by the brain? (easy to demonstrate). I will introduce you to synapses and describe classical and novel approaches to understand the brain (or at least better describe it). Importantly, I will emphasize how essential it is to study the brain at different levels of resolution and by applying an interdisciplinary approach. I promise to pose more questions than answers…


Michal Linial is a Professor of Biochemistry, The Hebrew University, Jerusalem, Israel and a Director of the Israel Institute for Advanced Studies.ML had published over 150 scientific papers and abstracts on diverse topics in molecular biology, cellular biology, bioinformatics, neuroscience the integration of tools to improve knowledge extractions. M. Linial has an experimental and computational laboratory. M.L is the leader and the founder of the first established educational program in Israel for Computer Science and Life Science (from 1999) for Undergraduate-Graduate studies. Her expertise in the synapse let to the study of protein families, protein-protein interactions with a global view on protein networks and their regulation. Molecular biology, cell biology and biochemical methods are applied in all research initiated in her laboratory. She and her laboratory are developing new computational and technological tools for large-scale cell biological research M. Linial and her colleagues apply MS based and genomics (DNA Chip) approaches for studying changes in neuronal development, and disease oriented research. She published over 180 scientific papers including book chapters and numerous reviews. The solid informatics approaches are used for large database storage and constant updating of several systems in view of classification, validation and functional predictions. M.L. and her students has been an active participant in NIH structural genomics initiatives and she participated in Structural Genomics effort Task for target selections. She and her colleagues have created several global classification systems that are used by the biomedical and biology communities. Most notably are the ProtoNet, EVEREST PANDORA, miRror,-Suite, ClanTox and more. All those developed web systems are provided as an open source for investigators.

Explore or Exploit? Reflections on an Ancient Dilemma in the Age of the Web – Robert Kleinberg

Robert Kleinberg, Cornell | Wednesday, August 6, 2014 | Video


Learning and decision-making problems often boil down to a balancing act between exploring new possibilities and exploiting the best known one. For more than fifty years, the multi-armed bandit problem has been the predominant theoretical model for investigating these issues. The emergence of the Web as a platform for sequential experimentation at a massive scale is leading to shifts in our understanding of this fundamental problem as we confront new challenges and opportunities. I will present two recent pieces of work addressing these challenges. The first concerns the misalignment of incentives in systems, such as online product reviews and citizen science platforms, that depend on a large population of users to explore a space of options. The second concerns situations in which the learner’s actions consume one or more limited-supply resources, as when a ticket seller experiments with prices for an event with limited seating.


Robert Kleinberg is an Associate Professor of Computer Science at Cornell University. His research studies the design and analysis of algorithms, and their relations to economics, learning theory, and networks. Prior to receiving his doctorate from MIT in 2005, Kleinberg spent three years at Akamai Technologies, where he assisted in designing the world’s largest Internet Content Delivery Network. He is the recipient of a Microsoft Research New Faculty Fellowship, an Alfred P. Sloan Foundation Fellowship, and an NSF CAREER Award.

Visual Nearest Neighbor Search – Shai Avidan

Shai Avidan, Tel-Aviv University | Wednesday, July 30, 2014 | Video


Template Matching finds the best match in an image to a given template and this is used in a variety of computer vision applications. I will discuss several extensions to Template Matching. First, dealing with the case where we have millions of templates that we must match at once, second dealing with the case of RGBD images, where depth information is available and finally, presenting a fast algorithm for template matching under 2D affine transformations with global approximation guarantees.

Joint work with Simon Korman, Yaron Eshet, Eyal Ofek, Gilad Tsur and Daniel Reichman.


Shai Avidan is an Associate Professor at the School of Electrical Engineering at Tel-Aviv University, Israel. He earned his PhD at the Hebrew University, Jerusalem, Israel, in 1999. Later, he was a Postdoctoral Researcher at Microsoft Research, a Project Leader at MobilEye, a startup company developing camera based driver assisted systems, a Research Scientist at Mitsubishi Electric Research Labs (MERL), and a Senior Researcher at Adobe. He published extensively in the fields of object tracking in video and 3-D object modeling from images. Recently, he has been working on Computational Photography. Dr. Avidan is an Associate Editor of PAMI and was on the program committee of multiple conferences and workshops in the fields of Computer Vision and Computer Graphics.

A Grand Gender Convergence: Its Last Chapter – Claudia Goldin

Claudia Goldin, Harvard | Wednesday, July 23, 2014 | Video


The converging roles of men and women are among the grandest advances in society and the economy in the last century. These aspects of the grand gender convergence are figurative chapters in a history of gender roles. But what must the “last” chapter contain for there to be equality in the labor market? The answer may come as a surprise. The solution does not (necessarily) have to involve government intervention and it need not make men more responsible in the home (although that wouldn’t hurt). But it must involve changes in the labor market, in particular how jobs are structured and remunerated to enhance temporal flexibility. The gender gap in pay would be considerably reduced and might vanish altogether if firms did not have an incentive to disproportionately reward individuals who labored long hours and worked particular hours. Such change has taken off in various sectors, such as technology, science and health, but is less apparent in the corporate, financial and legal worlds.


Claudia Goldin is the Henry Lee Professor of Economics at Harvard University and director of the NBER’s Development of the American Economy program. Goldin is an economic historian and a labor economist. Her research has covered a wide array of topics, such as slavery, emancipation, the post-bellum south, women in the economy, the economic impact of war, immigration, New Deal policies, inequality, technological change, and education. Most of her research interprets the present through the lens of the past and explores the origins of current issues of concern. In the past several years her work has concerned the rise of mass education in the United States and its impact on economic growth and wage inequality. More recently she has focused her attention on college women’s achievement of career and family.

She is the author and editor of several books, among them Understanding the Gender Gap: An Economic History of American Women (Oxford 1990), The Regulated Economy: A Historical Approach to Political Economy (with G. Libecap; University of Chicago Press 1994), The Defining Moment: The Great Depression and the American Economy in the Twentieth Century (with M. Bordo and E. White; University of Chicago Press 1998), and Corruption and Reform: Lesson’s from America’s Economic History (with E. Glaeser; Chicago 2006). Her most recent book is The Race between Education and Technology (with L. Katz; The Belknap Press, 2008), winner of the 2008 R.R. Hawkins Award for the most outstanding scholarly work in all disciplines of the arts and sciences.Goldin is best known for her historical work on women in the U.S. economy. Her most recent papers in that area have concerned the history of women’s quest for career and family, coeducation in higher education, the impact of the “pill” on women’s career and marriage decisions, women’s surnames after marriage as a social indicator, and the reasons why women are now the majority of undergraduates. She has recently embarked on a wide ranging project on the family and career transitions of male and female graduates of selective universities from the late 1960s to the present.Goldin is the current president of the American Economic Association. In 2007 Goldin was elected a member of the National Academy of Sciences and was the Gilman Fellow of the American Academy of Political and Social Science. She is a fellow of the American Academy of Arts and Sciences, the Society of Labor Economists (SOLE), the Econometric Society, and the Cliometric Society. In 2009 SOLE awarded Goldin the Mincer Prize for life-time contributions to the field of labor economics. Goldin completed her term as the President of the Economic History Association in 2000. In 1991 she was elected Vice President of the American Economic Association. From 1984 to 1988 she was editor of the Journal of Economic History and is currently an associate editor of the Quarterly Journal of Economics and a member of various editorial boards. She is the recipient of various teaching awards. Goldin received her B.A. from Cornell University and her Ph.D. from the University of Chicago.

Why Not Be Evil? The Costs and Benefits of Corporate Social Responsibility – Siva Vaidhyanathan

Siva Vaidhyanathan, University of Virginia | Wednesday, July 9, 2014


Corporate Social Responsibility (CSR) and its Silicon Valley cousin, Social Entrepreneurship, have a rich but recent history. This talk will briefly explore the roots of these schools of thought and practice and examine their rise through business-school curricula and scholarship in the late 20th Century. Why did they come about when they came about? What are their effects on the world? Do they affect consumer behavior and investor behavior? And to what ends? Most seriously, does the identification of a company with particular values or social goals have the effect of depoliticizing an otherwise democratic republic?


Siva Vaidhyanathan is the Robertson Professor of Media Studies at the University of Virginia and the author, most recently, of The Googlization of Everything — and Why We Should Worry (University of California Press, 2011)

Rethinking Machine Learning In The 21St Century: From Optimization To Equilibration – Sridhar Mahadevan

Sridhar Mahadevan, UMASS Amherst | Wednesday, June 11, 2014 | Video


The past two decades has seen machine learning (ML) transformed from an academic curiosity to a multi-billion dollar industry, and a centerpiece of our economic, social, scientific, and security infrastructure. Much work in machine learning has drawn on research in optimization, motivated by large-scale applications requiring analysis of massive high-dimensional data. In this talk, I’ll argue that the growing importance of networked data environments, from the Internet to cloud computing, requires a fundamental rethinking of our basic analytic tools. My thesis will be that ML needs to shift from its current focus on optimization to equilibration, from modeling the world as uncertain, but stationary and benign, to one where the world is non-stationary, competitive, and potentially malicious. Adapting to this new world will require developing new ML frameworks and algorithms. My talk will introduce one such framework — equilibration using variational inequalities and projected dynamical systems —which not only generalizes optimization, but is better suited to the distributed networked cloud-oriented future that ML faces. To explain this paradigm change, I’ll begin by summarizing the au courant optimization-based approach to ML using recent research in the Autonomous Learning Laboratory. I will then present an equilibration-based framework using variational inequalities and projected dynamical systems, which originated in mathematics for solving partial differential equations in physics, but has been since been widely applied in its finite-dimensional formulation to network equilibrium problems in economics, transportation, and other areas. I’ll describe a range of algorithms for solving variational inequalities, showing their scope allows ML to extend beyond optimization, to finding game-theoretic equilibria, solving complementarity problems, and many other areas.


Professor Sridhar Mahadevan directs the Graduate Program at the School of Computer Science at the University of Massachusetts, Amherst. He is a co-director of the Autonomous Learning Laboratory, one of the oldest academic research centers for machine learning in the US, which has graduated more than 30 doctoral students in its three decade history, and includes 3 AAAI fellows among its alumni. The lab currently includes 14 PhD students, who work in a variety of areas in machine learning, including equilibration algorithms, optimization, reinforcement learning, and unsupervised learning.

Do Neighborhoods Matter for Disadvantaged Families? Long-Term Evidence from the Moving to Opportunity Experiment – Larry Katz

Larry Katz, Harvard | Wednesday, May 21, 2014 | Video


We examine long-term neighborhood effects on low-income families using data from the Moving to Opportunity (MTO) randomized housing-mobility experiment, which offered some public-housing families but not others the chance to move to less-disadvantaged neighborhoods. MTO succeed in moving families to lower-poverty and safer residential neighborhoods, but MTO moves did not substantially improve the quality of schools attended by the children. We show that 10-15 years after baseline, MTO improves adult physical and mental health, has no detectable effect on economic outcomes or youth schooling or physical health, and mixed results by gender on other youth outcomes, with girls doing better on some measures and boys doing worse. Despite the somewhat mixed pattern of impacts on traditional behavioral outcomes, MTO moves substantially improve adult subjective well-being. And when opportunities to move with housing vouchers lead to better schools for the children, such moves do have long-run positive impacts on youth education and reduce youth risky behaviors.


Lawrence F. Katz is the Elisabeth Allison Professor of Economics at Harvard University and a Research Associate of the National Bureau of Economic Research. His research focuses on issues in labor economics and the economics of social problems. He is the author (with Claudia Goldin) of The Race between Education and Technology (Harvard University Press, 2008), a history of U.S. economic inequality and the roles of technological change and the pace of educational advance in affecting the wage structure.

Katz also has been studying the impacts of neighborhood poverty on low-income families as the principal investigator of the long-term evaluation of the Moving to Opportunity program, a randomized housing mobility experiment. And Katz is working with Claudia Goldin on a major project studying the historical evolution of career and family choices and outcomes for U.S. college men and women. His past research has explored a wide range of topics including U.S. and comparative wage inequality trends, educational wage differentials and the labor market returns to education, the impact of globalization and technological change on the labor market, the economics of immigration, unemployment and unemployment insurance, regional labor markets, the evaluation of labor market programs, the problems of low-income neighborhoods, and the social and economic consequences of the birth control pill.

Professor Katz has been editor of the Quarterly Journal of Economics since 1991 and served as the Chief Economist of the U.S. Department of Labor for 1993 and 1994. He is the co-Scientific Director of J-PAL North America, current President of the Society of Labor Economists, and has been elected a fellow of the National Academy of Sciences, American Academy of Arts and Sciences, the Econometric Society, and the Society of Labor Economists. Katz serves on the Panel of Economic Advisers of the Congressional Budget Office as well as on the Boards of the Russell Sage Foundation and the Manpower Demonstration Research Corporation. He graduated from the University of California at Berkeley in 1981 and earned his Ph.D. in Economics from the Massachusetts Institute of Technology in 1985.

Principled Approaches for Learning Latent Variable Models – Anima Anandkumar

Anima Anandkumar, UC Irvine | Wednesday, May 14, 2014 | Video


In any learning task, it is natural to incorporate latent or hidden variables which are not directly observed. For instance, in a social network, we can observe interactions among the actors, but not their hidden interests/intents, in gene networks, we can measure gene expression levels but not the detailed regulatory mechanisms, and so on. I will present a broad framework for unsupervised learning of latent variable models, addressing both statistical and computational concerns. We show that higher order relationships among observed variables have a low rank representation under natural statistical constraints such as conditional-independence relationships. We also present efficient computational methods for finding these low rank representations. These findings have implications in a number of settings such as finding hidden communities in networks, discovering topics in text documents and learning about gene regulation in computational biology. I will also present principled approaches for learning overcomplete models, where the latent dimensionality can be much larger than the observed dimensionality, under natural sparsity constraints. This has implications in a number of applications such as sparse coding and feature learning.


Anima Anandkumar is a faculty at the EECS Dept. at U.C. Irvine since August 2010. Her research interests are in the area of large-scale machine learning and high-dimensional statistics. She received her B.Tech in Electrical Engineering from IIT Madras in 2004 and her PhD from Cornell University in 2009. She was a postdoctoral researcher at the Stochastic Systems Group at MIT between 2009- 2010. She is the recipient of the Alfred P. Sloan Fellowship, Microsoft Faculty Fellowship, ARO Young Investigator Award, NSF CAREER Award, IBM Fran Allen PhD fellowship, thesis award from ACM SIGMETRICS society, and paper awards from the ACM SIGMETRICS and IEEE Signal Processing societies.

Economies of Visibility: Girl Empowerment Organizations and the Market for Empowerment – Sarah Banet-Weiser

Sarah Banet-Weiser, USC Annenberg’s School of Communication | Wednesday, April 30, 201


In the past two decades, the invocation of “girl power” as an increasingly normative discourse to describe young girls and women in their everyday practices has been met with both excitement and challenge. However, while many have theorized how the “girl” in girl power is a racially and class specific girl, one that has economic and cultural privilege to access power, the “power” in girl power still needs rigorous theorization. In this talk, I examine what the “power” of girl power means in the current moment, arguing that for the most part, this form of power is legible within an economy of media visibility, where media incessantly look at and invite us to look at girls. More specifically, I examine the construction of a market within the contemporary economy of visibility: the market for empowerment. Looking at girl empowerment organizations, I analyze this market in both a US and international development context, and argue that it works to consolidate a specific kind of empowerment that is personal and individual.


Sarah Banet-Weiser is Professor of Communication at the Annenberg School of Communication and Journalism and in the Department of American Studies and Ethnicity at the University of Southern California. She is the author of The Most Beautiful Girl in the World: Beauty Pageants and National Identity (1999), Kids Rule! Nickelodeon and Consumer Citizenship (2007), and Authentic™: The Politics of Ambivalence in a Brand Culture (winner of the Outstanding Book Award at the International Communication Association). She is the co-editor of Cable Visions: Television Beyond Broadcasting and Commodity Activism: Cultural Resistance in Neoliberal Times. She edited the NYU press book series Critical Cultural Communication until 2012, and is currently the editor of American Quarterly.

Those Of You Who Need a Little More Time – Jonathan Sterne

Jonathan Sterne, McGill | Wednesday, April 16, 201


This talk examines the lesser-known work and legacy of Dennis Gabor. Gabor was a physicist famous for inventing holography. But he also applied quantum theory to sound, and in so doing offered an important corrective to prevailing interpretations of wave theories of sound derived from Joseph Fourier’s work. To prove his point, Gabor built a device called the “kinematic frequency compressor,” which could time-stretch or pitch-shift audio independently of the other operation, a feat previously considered impossible in the analog domain. After considering the machine, I trace its technical and cultural descendants in advertising, cinema, avant-garde music, and today in the world’s most popular audio software, Ableton Live.


Jonathan Sterne teaches in the Department of Art History and Communication Studies and the History and Philosophy of Science Program at McGill University. He is author of MP3: The Meaning of a Format (Duke 2012), The Audible Past: Cultural Origins of Sound Reproduction (Duke, 2003); and numerous articles on media, technologies and the politics of culture. He is also editor of The Sound Studies Reader (Routledge, 2012). His new projects consider instruments and instrumentalities; histories of signal processing; and the intersections of disability, technology and perception. Visit his website at

Deceptive Products – Botond Koszegi

Botond Koszegi, Central European University | Wednesday, April 2, 201


A literature in behavioral economics documents that in a number of retail markets, some consumers misunderstand key fees or other central product features, and many argue that this leads firms to offer contracts and products that take advantage of such naive consumers. This talk will give an overview of some theoretical research on the market for deceptive products. Questions might include (i) what kinds of contracts will be offered in the presence of naive consumers; (ii) how naive and sophisticated consumers affect each other in the market; (iii) how firms attempt to discriminate between naive and sophisticated consumers, and how this affects economic welfare; (iv) whether and when firms have an incentive to “come clean” regarding their products; and (v) what kinds of products will be sold in a deceptive way.

Based on joint work with Paul Heidhues and Takeshi Murooka


Botond Koszegi is Professor at the Department of Economics at Central European University in Budapest, Hungary, since August 1, 2012. He was previously Professor of Economics at the University of California at Berkeley, and has held visiting positions at Massachusetts Institute of Technology, Cambridge, MA, and CEU. He earned his BA in mathematics from Harvard University in 1996, and his Ph.D. in economics from the Massachusetts Institute of Technology in 2000. His research interests are primarily in the theoretical foundations of behavioral economics. He has produced research on self-control problems and the consumption of harmful products, self-image and anticipatory utility, reference-dependent preferences and loss aversion, and focusing and attention. Recently, he has been studying how firms respond to consumers’ psychological tendencies, especially in the pricing of products and the design of credit and other financial contracts.

Mechanism Design for Data Science – Jason Hartline

Jason Hartline, Northwestern | Wednesday, March 19, 2014 | Video


The promise of data science is that system data can be analyzed and its understanding can be used to improve the system (i.e., to obtain good outcomes). For this promise to be realized, the necessary understanding must be inferable from the data. Whether or not this understanding is inferable often depends on the system itself. Therefore, the system needs to be designed to both obtain good outcomes and to admit good inference. This talk will explore this issue in a mechanism design context where the designer would like use past bid data to adapt an auction mechanism to optimize revenue. Data analysis is necessary for revenue optimization in auctions, but revenue optimization is at odds with good inference. The revenue-optimal auction for selling an item is typically parameterized by a reserve price, and the appropriate reserve price depends on how much the bidders are willing to pay. This willingness to pay could be potentially be learned by inference, but a reserve price precludes learning anything about willingness-to-pay of bidders who are not willing to pay the reserve price. The auctioneer could never learn that lowering the reserve price would give a higher revenue (even if it would). To address this impossibility, the auctioneer could sacrifice revenue-optimality in the initial auction to obtain better inference properties so that the auction’s parameters can be adapted to changing preferences in the future. In this talk, I will develop a theory for optimal auction design subject to good inference.


Prof. Hartline is on sabbatical at Harvard Economics and Computer Science Departments for the 2014 calendar year (January 2014-December 2014).

Prof. Hartline’s current research interests lie in the intersection of the fields of theoretical computer science, game theory, and economics. With the Internet developing as the single most important arena for resource sharing among parties with diverse and selfish interests, traditional algorithmic and distributed systems approaches are insufficient. Instead, in protocols for the Internet, game-theoretic and economic issues must be considered. A fundamental research endeavor in this new field is the design and analysis of auction mechanisms and pricing algorithms.

Dr. Hartline joined the EECS department (and MEDS, by courtesy) in January of 2008. He was a researcher at Microsoft Research, Silicon Valley from 2004 to 2007, where his research covered foundational topic of algorithmic mechanism design and applications to auctions for sponsored search. He was an active researcher in the San Francisco bay area algorithmic game theory community and was a founding organizer of the Bay Algorithmic Game Theory Symposium. In 2003, he held a postdoctoral research fellowship at the Aladdin Center at Carnegie Mellon University. He received his Ph.D. in Computer Science from the University of Washington in 2003 with advisor Anna Karlin and B.S.s in Computer Science and Electrical Engineering from Cornell University in 1997.

An Experiment in Hiring Discrimination Via Online Social Networks – Alessandro Acquisti

Alessandro Acquisti, CMU | Wednesday, Feb 26, 201


Surveys of U.S. employers suggest that numerous firms seek information about job applicants online. However, little is known about how this information gathering influences employers’ hiring behavior. We present results from two complementary randomized experiments (a field experiment and an online experiment) on the impact of online information on U.S. firms’ hiring behavior. We manipulate candidates’ personal information that is protected under either federal laws or some state laws, and may be risky for employers to enquire about during interviews, but which may be inferred from applicants’ online social media profiles. In the field experiment, we test responses of over 4,000 U.S. employers to a Muslim candidate relative to a Christian candidate, and to a gay candidate relative to a straight candidate. We supplement the field experiment with a randomized, survey-based online experiment with over 1,000 subjects (including subjects with previous human resources experience) testing the effects of the manipulated online information on hypothetical hiring decisions and perceptions of employability. The results of the field experiment suggest that a minority of U.S. firms likely searched online for the candidates’ information. Hence, the overall effect of the experimental manipulations on interview invitations is small and not statistically significant. However, in the field experiment, we find evidence of discrimination linked to political party affiliation. Following the Gallup Organization’s segmentation of U.S. states by political ideology, we use results from the 2012 presidential election and find evidence of discrimination against the Muslim candidate compared to the Christian candidate among employers in more Romney-leaning states and counties. These results are robust to controlling for firm characteristics, state fixed effects, and a host of county-level variables. We find no evidence of discrimination against the gay candidate relative to the straight candidate. Results from the online experiment are consistent with those from the field experiment: we find more evidence of bias among subjects more likely to self-report more political conservative party affiliation.


Alessandro Acquisti is a professor of information technology and public policy at the Heinz College, Carnegie Mellon University (CMU) and the co-director of CMU Center for Behavioral and Decision Research. He has held visiting positions at the Universities of Rome, Paris, and Freiburg (visiting professor); Harvard University (visiting scholar); University of Chicago (visiting fellow); Microsoft Research (visiting researcher); and Google (visiting scientist). Alessandro investigates economic, policy, and technological issues surrounding privacy. His studies have spearheaded the application of behavioral economics to the analysis of privacy and information security decision making, and the analysis of privacy risks and disclosure behavior in online social networks. Alessandro has been the recipient of the PET Award for Outstanding Research in Privacy Enhancing Technologies, the IBM Best Academic Privacy Faculty Award, multiple Best Paper awards, and the Heinz College School of Information’s Teaching Excellence Award. He has testified before the U.S. Senate and House committees on issues related to privacy policy and consumer behavior, and was a TED Global 2013 speaker. Alessandro’s findings have been featured in national and international media outlets, including the Economist, the New York Times, the Wall Street Journal, the Washington Post, the Financial Times,, NPR, CNN, and CBS 60 Minutes. His 2009 study on the predictability of Social Security numbers was featured in the “Year in Ideas” issue of the New York Times Magazine. Alessandro holds a PhD from UC Berkeley, and Master degrees from UC Berkeley, the London School of Economics, and Trinity College Dublin. He has been a member of the National Academies’ Committee on public response to alerts and warnings using social media.

Economic Models as Analogies – Larry Samuelson

Larry Samuelson, Yale | Wednesday, December 18, 201


People often wonder why economists analyze models whose assumptions are known to be false, while economists feel that they learn a great deal from such exercises. We suggest that part of the knowledge generated by academic economists is case-based rather than rule-based. That is, instead of offering general rules or theories that should be contrasted with data, economists often analyze models that are “theoretical cases”, which help understand economic problems by drawing analogies between the model and the problem. According to this view, economic models, empirical data, experimental results and other sources of knowledge are all on equal footing, that is, they all provide cases to which a given problem can be compared. We offer complexity arguments that explain why case-based reasoning may sometimes be the method of choice and why economists prefer simple cases.

Joint work with Itzhak Gilboa, Andrew Postlewaite, and David Schmeidler


Samuelson is a Fellow of the Econometric Society and a Fellow of the American Academy of Arts and Sciences. He has been a Co-editor of Econometrica and is currently a Co-editor of the American Economic Review. His research spans microeconomic theory and game theory.

Tools for Large Scale Public Engagement in Research – Krzysztof Gajos

Krzysztof Gajos, Harvard | Wednesday, December 4, 201


Non-scientists have long been contributing to research: by gathering observations on plant and animal behavior, by gazing at the sky through private amateur telescopes, or by participating in psychology experiments. The Internet has created entirely new opportunities for enabling public participation in research, both in terms of the scale of public participation and the kinds of activities that the non-professional scientists can perform in support of scientific inquiry. Yet, inclusion of the broader publics in one’s research program remains an exception rather than a norm, presumably because of concerns related to technical infrastructure, recruitment, and reliability of contributions.I will highlight two strands of research in my group that contribute toward wider involvement of broader publics in research.

In the first strand, we have specifically focused on methods for studying human motor performance on computer input tasks. We have developed and validated mechanisms for collecting lab-quality data in three settings: 1. unobtrusively in situ from observations of a user’s natural interactions with a computer; 2. on Amazon Mechanical Turk; 3. with unpaid online volunteers through our Lab in the Wild platform. Our recent study with 500,000 participants allowed us to replicate several past results and also to conduct new analyses that were not possible before. For example, we provided fine grained estimates of when in life basic abilities (such as cognitive processing speed, fine motor control, and gross motor control) peak.

In the second strand, we focused on developing procedures to enable non-experts to perform expert-level analytical tasks accurately and at scale. Specifically, we have developed PlateMate, a system for crowdsourcing nutritional analysis from food photographs. In an ongoing project, we are studying the behavioral and nutritional factors impacting preterm birth. A key technical enabler of this project is a mechanism, based on our PlateMate system, for scalable nutritional analysis, which will make it possible to track the nutritional intake of 400 pregnant women for several months each.


Krzysztof Z. Gajos is an associate professor of computer science at the Harvard School of Engineering and Applied Sciences. Krzysztof is primarily interested in intelligent interactive systems, an area that spans human-computer interaction, artificial intelligence, and applied machine learning. Krzysztof received his B.Sc. and M.Eng. degrees in Computer Science from MIT. Subsequently he was a research scientist at the MIT Artificial Intelligence Laboratory, where he managed The Intelligent Room Project. In 2008, he received his Ph.D. in Computer Science from the University of Washington in Seattle. Before coming to Harvard in September of 2009, he spent a year as a post-doctoral researcher in the Adaptive Systems and Interaction group at Microsoft Research.URL:

Understanding Audition Via Sound Synthesis – Josh McDermott

Josh McDermott, MIT | Wednesday, November 20, 2013


Humans infer many important things about the world from the sound pressure waveforms that enter the ears. In doing so we solve a number of difficult and intriguing computational problems. We recognize sound sources despite large variability in the waveforms they produce, extract behaviorally relevant attributes that are not explicit in the input to the ear, and do so even when sound sources are embedded in dense mixtures with other sounds. This talk will describe recent progress in understanding these remarkable auditory abilities. The work stems from the premise that a theory of the perception of some property should enable the synthesis of signals that appear to have that property. Sound synthesis can thus be used to test theories of perception and to explore representations of sound. I will describe several examples of this approach.


Josh McDermott is a perceptual scientist studying sound, hearing, and music in the Department of Brain and Cognitive Sciences at MIT. His research addresses human and machine audition using tools from experimental psychology, engineering, and neuroscience. He is particularly interested in using the gap between human and machine competence to both better understand biological hearing and design better algorithms for analyzing sound.

McDermott obtained a BA in Brain and Cognitive Science from Harvard, an MPhil in Computational Neuroscience from University College London, a PhD in Brain and Cognitive Science from MIT, and postdoctoral training in psychoacoustics at the University of Minnesota and in computational neuroscience at NYU. He is the recipient of a Marshall Scholarship, a National Defense Science and Engineering fellowship, and a James S. McDonnell Foundation Scholar Award. He is currently an Assistant Professor in the Department of Brain and Cognitive Sciences at MIT.

Graphical approaches to Biological Problems – Ernest Fraenkel

Ernest Fraenkel, MIT | Wednesday, November 6, 2013 | Video


Biology has been transformed by new technologies that provide detailed descriptions of the molecular changes that occur in diseases. However, it is difficult to use these data to reveal new therapeutic insights for several reasons. Despite their power, each of these methods still only captures a small fraction of the cellular response. Moreover, when different assays are applied to the same problem, they provide apparently conflicting answers. I will show that network modeling reveals the underlying consistency of the data by identifying small, functionally coherent pathways linking the disparate observations. We have used these methods to analyze how oncogenic mutations alter signaling and transcription and to prioritize experiments aimed at discovering therapeutic targets.


Ernest Fraenkel was first introduced to computational biology in high school when the field did not yet have a name. His early experiences with Professor Cyrus Levinthal of Columbia University taught him that biological insights often come from unexpected disciplines. After graduating summa cum laude from Harvard College in Chemistry and Physics he obtained his Ph.D. at MIT in the department of Biology and did post-doctoral work at Harvard. As the field of Systems Biology began to emerge, he established a research group in this area at the Whitehead Institute and then moved to the Department of Biological Engineering at the Massachusetts Institute of Technology. His research group takes a multi-disciplinary approach involving tightly connected computational and experimental methods to uncover the molecular pathways that are altered in cancer, neurodegenerative diseases, and diabetes.

Social Norms and the Impact of Laws – Matt Jackson

Matt Jackson, Stanford | Wednesday, September 18, 2013


We examine the impact of laws in a model of social norms. Agents each choose a level of behavior (e.g., a speed of driving, an amount of corruption, etc.). Agents choose behaviors not only based on their personal preference but also based on a preference to match or conform to the behaviors of other agents with whom they interact. A law caps the level of behavior and a law-abiding agent may whistle-blow on an agent who is breaking the law: correcting the behavior of the latter and making him or her pay a fine. The impact of a law is endogenous to the social norm (equilibrium of behavior) and as such laws can have nonmonotone effects: a strict law may be broken more frequently than an lax one. Moreover, law-breakers may choose more extreme behavior as a law becomes stricter. Historical behavior can influence the impact of a law: exactly the same law can have drastically different impacts in two different societies depending on past social norms.


Matthew O. Jackson is the Eberle Professor of Economics at Stanford University and an external faculty member of the Santa Fe Institute and a fellow of CIFAR. Jackson’s research interests include game theory, microeconomic theory, and the study of social and economic networks, including diffusion, learning, and network formation. He was at Northwestern and Caltech before joining Stanford, and has a PhD from Stanford and BA from Princeton. Jackson is a Fellow of the Econometric Society and the American Academy of Arts and Sciences, and former Guggenheim Fellow.

Crowdsourcing Audio Production Interfaces – Bryan Pardo

Bryan Pardo, Northwestern | Wednesday, September 11, 2013 | Video


Potential users of audio production software, such as audio equalizers, may be discouraged by the complexity of the interface and a lack of clear affordances in typical interfaces. We seek to simplify interfaces for task such as audio production (e.g. mastering a music album with ProTools), audio tools (e.g. equalizers) and related consumer devices (e.g. hearing aids). Our approach is to use an evaluative paradigm (“I like this sound better than that sound”) with the use of descriptive language (e.g. “Make the violin sound ‘warmer.’”). To achieve this goal, a system must be able to tell whether the stated goal is appropriate for the selected tool (e.g. making the violin “warmer” with a panning tool does not make sense). If the goal is appropriate for the tool, it must know what actions need to be taken (e.g. add some reverberation). Further, the tool should not impose a vocabulary on users, but rather understand the vocabulary users prefer. In this talk, Bryan Pardo describes, iQ, an equalizer that uses an evaluative control paradigm and SocialEQ, a web-based project to crowdsource a vocabulary of actionable audio descriptors.


Bryan Pardo, head of the Northwestern University Interactive Audio Lab, is an associate professor in the Northwestern University Department of Electrical Engineering and Computer Science. Prof. Pardo received a M. Mus. in Jazz Studies in 2001 and a Ph.D. in Computer Science in 2005, both from the University of Michigan. He has authored over 70 peer-reviewed publications. He has developed speech analysis software for the Speech and Hearing department of the Ohio State University, statistical software for SPSS and worked as a machine learning researcher for General Dynamics. While finishing his doctorate, he taught in the Music Department of Madonna University. When he’s not programming, writing or teaching, he performs throughout the United States on saxophone and clarinet at venues such as Albion College, the Chicago Cultural Center, the Detroit Concert of Colors, Bloomington Indiana’s Lotus Festival and Tucson’s Rialto Theatre.

Seeing the invisible; Predicting the unexpected – Michal Irani

Michal Irani, Weizmann | Wednesday, September 4, 201


In this talk I will show how complex visual inference tasks can be performed, with no prior examples, by exploiting internal redundancy within visual data. Comparing and integrating local pieces of visual information gives rise to complex notions of visual similarity and to a general “Inference by Composition” approach. This allows to infer about the likelihood of new visual data that was never seen before, and make inferences about complex static and dynamic visual information without any prior examples. I will demonstrate the power of this approach to several example problems (as time permits):

  1. Detecting complex objects and actions.
  2. Prediction of missing visual information.
  3. Inferring the “likelihood” of “never-before-seen” visual data.
  4. Detecting the “irregular” and “unexpected”
  5. Spatial super-resolution (from a single image) & Temporal super-resolution (from a single video).
  6. Generating visual summaries (of images and videos)
  7. Segmentation of complex visual data.


Michal Irani is a Professor at the Weizmann Institute of Science, in the Department of Computer Science and Applied Mathematics. She received a B.Sc. degree in Mathematics and Computer Science from the Hebrew University of Jerusalem in 1985, and M.Sc. and Ph.D. degrees in Computer Science from the same institution in 1989 and 1994, respectively. From 1993 to 1996, she was a member of the technical staff of the Vision Technologies Laboratory at the David Sarnoff Research Center (Princeton, New Jersey, USA). She joined the Weizmann Institute at 1997. Michal’s research interests center around computer vision, image processing, and video information analysis. Michal’s prizes and honors include the David Sarnoff Research Center Technical Achievement Award (1994), the Yigal Allon three-year Fellowship for Outstanding Young Scientists (1998), and the Morris L. Levinson Prize in Mathematics (2003). At the European Conference on Computer Vision, she received awards for Best Paper in 2000 and in 2002, and was awarded an Honorable Mention for the Marr Prize at the IEEE International Conference on Computer Vision in 2001 and in 2005.

Differential Privacy: Theoretical and Practical Challenges – Salil Vadhan

Salil Vadhan, Harvard | Wednesday, August 14, 201


Differential Privacy is framework for enabling the analysis of privacy-sensitive datasets while ensuring that individual-specific information is not revealed. The concept was developed in a body of work in theoretical computer science starting about a decade ago, largely coming from Microsoft Research. It is now flourishing as an area of theory research, with deep connections to many other topics in theoretical computer science. At the same time, its potential for addressing pressing privacy problems in a variety of domains has attracted the interest of scholars from many other areas, including statistics, databases, medical informatics, law, social science, computer security and programming languages.

In this talk, I will give a general introduction to differential privacy, and discuss some of the theoretical and practical challenges for future work in this area. I will also describe a large, multidisciplinary research project at Harvard, called “Privacy Tools for Sharing Research Data,” in which we are working on some of these challenges as well as others associated with the collection, analysis, and sharing of personal data for research in social science and other fields.


Salil Vadhan is the Vicky Joseph Professor of Computer Science and Applied Mathematics at the School of Engineering & Applied Sciences at Harvard University. He is a member of the Theory of Computation research group. His research areas include computational complexity, cryptography, randomness in computation, and data privacy.

Technologies of Choice? – ICTs, development and the capabilities approach – Dorothea Kleine

Dorothea Kleine, University of London | Wednesday, July 31, 2013 | Video


ICT for development (ICT4D) scholars claim that the internet, radio and mobile phones can support development. Yet the dominant paradigm of development as economic growth is too limiting to understand the full potential of these technologies. One key rival to such econocentric understandings is Amartya Sen’s capabilities approach to development – focusing on a pluralistic understanding of people’s values and the lives they want to lead. In her book, Technologies of Choice? (MIT Press 2013), Dorothea Kleine translates Sen’s approach into policy analysis and ethnographic work on technology adaptation. She shows how technologies are not neutral, but imbued with values that may or may not coincide with the values of users. The case study analyses Chile’s pioneering ICT policies in the areas of public access, digital literacy, and online procurement and the sobering reality of one of the most marginalised communities in the country where these policies play out. The book shows how both neoliberal and egalitarian ideologies are written into technologies as they permeate the everyday lives and livelihoods of women and men in the town. Technologies of Choice? examines the relationship between ICTs, choice, and development. It argues for a people-centred view of development that has individual and collective choice at its heart.


Dorothea Kleine is Senior Lecturer in Human Geography and Director of the interdisciplinary ICT4D Centre at Royal Holloway, University of London ( In 2013 the Centre was named among the top 10 global think tanks in science and technology (U of Penn survey of experts) and has a highly recognized PhD and Masters program in ICT for development. Dorothea’s work focuses on the relationship between notions of “development”, choice and individual agency, sustainability, gender and technology. She has published widely on these subjects, and has worked as an advisor to UNICEF, UNEP, EUAid, DFID, GIZ and to NGOs. The Centre runs various collaborative research projects with international agencies and private sector partners.

The Cryptographic Lens – Shafi Goldwasser

Shafi Goldwasser, MIT | Wednesday, July 17, 201


Going beyond the basic challenge of private communication, in the last 35 years, cryptography has become the general study of correctness and privacy of computation in the presence of a computationally bounded adversary, and as such has changed how we think of proofs, reductions, randomness, secrets, and information. In this talk I will discuss some beautiful developments in the theory of computing through this cryptographic lens, and the role cryptography can play in the next successful shift from local to global computation.


Goldwasser is the RSA Professor of Electrical Engineering and Computer Science at MIT and a professor of computer science and applied mathematics at the Weizmann Institute of Science. Goldwasser received a BS (1979) in applied mathematics from CMU and PhD (1984) in computer science from UC Berkeley. Goldwasser is the 2012 recipient of the ACM Turing Award.

Does the Classic Microfinance Model Discourage Entrepreneurship Among the Poor? Experimental Evidence from India – Erica Field

Erica Field, Duke Wednesday, July 10, 2013 | Video


Do the repayment requirements of the classic microfinance contract inhibit investment in high-return but illiquid business opportunities among the poor? Using a field experiment, we compare the classic contract which requires that repayment begin immediately after loan disbursement to a contract that includes ta two-month grace period. The provision of a grace period increased short-run business investment and long-run profits but also default rates. The results, thus, indicate that debt contracts that require early repayment discourage illiquid risky investment and thereby limit the potential impact of microfinance on microenterprise growth and household poverty.


Erica M. Field joined the Duke faculty as an associate professor in 2011. She is also a faculty research fellow at the National Bureau of Economic Research. Professor Field received her Ph.D. and M.A. in economics from Princeton University in 2003 and her B.A. in economics and Latin American studies from Vassar College in 1996. Since receiving her doctorate, she has worked at Princeton, Stanford, and most recently Harvard, where she was a professor for six years before coming to Duke.

Machine Learning for Complex Social Processes – Hanna Wallach

Hanna Wallach, UMass Amherst | Wednesday, July 3, 201


From the activities of the US Patent Office or the National Institutes of Health to communications between scientists or political legislators, complex social processes—groups of people interacting with each other in order to achieve specific and sometimes contradictory goals—underlie almost all human endeavor. In order draw thorough, data-driven conclusions about complex social processes, researchers and decision-makers need new quantitative tools for exploring, explaining, and making predictions using massive collections of interaction data. In this talk, I will discuss the development of machine learning methods for modeling interaction data. I will concentrate on exploratory analysis of communication networks — specifically, discovery and visualization of topic-specific subnetworks in email data sets. I will present a new Bayesian latent variable model of network structure and content and explain how this model can be used to analyze intra-governmental email networks.


In fall 2010, Hanna Wallach started as an assistant professor in the Department of Computer Science at the University of Massachusetts Amherst. She is one of five core faculty members involved in UMass’s new Computational Social Science Initiative. Prior to this, Hanna was a senior postdoctoral research associate, also at UMass, where she developed statistical machine learning techniques for analyzing complex data regarding communication and collaboration within scientific and technological innovation communities. Hanna’s Ph.D. work, undertaken at the University of Cambridge, introduced new methods for statistically modeling text using structured topic models—models that automatically infer semantic information from unstructured text and information about document structure, ranging from sentence structure to inter-document relationships. Hanna holds an M.Sc. from the University of Edinburgh, where she specialized in neural computing and learning from data, and was awarded the University of Edinburgh’s 2001/2002 prize for Best M.Sc. Student in Cognitive Science. Hanna received her B.A. from the University of Cambridge Computer Laboratory in 2001. Her undergraduate project, “Visual Representation of Computer-Aided Design Constraints,” won the award for the best computer science student in the 2001 U.K. Science Engineering and Technology Awards. In addition to her many papers on statistical machine learning techniques for analyzing structured and unstructured data, Hanna’s tutorial on conditional random fields is extremely widely cited and used in machine learning courses around the world. Her recent work (with Ryan Prescott Adams and Zoubin Ghahramani) on infinite belief networks won the best paper award at AISTATS 2010. As well as her research, Hanna works to promote and support women’s involvement in computing. In 2006, she co-founded an annual workshop for women in machine learning, in order to give female faculty, research scientists, postdoctoral researchers, and graduate students an opportunity to meet, exchange research ideas, and build mentoring and networking relationships. In her not-so-spare time, Hanna is a member of Pioneer Valley Roller Derby, where she is better known as Logistic Aggression.

Crowd Computing – Rob Miller

Rob Miller, MIT |  Wednesday, June 19, 2013 | Video


Crowd computing harnesses the power of people out in the web to do tasks that are hard for individual users or computers to do alone. Like cloud computing, crowd computing offers elastic, on-demand human resources that can drive new applications and new ways of thinking about technology. This talk will describe several prototype systems we have built, including:

  • Soylent, a Word plugin that crowdsources text editing tasks;
  • VizWiz, an app that helps blind people see using a crowd’s eyes;
  • Adrenaline, a camera shutter driven by crowd perception;
  • Caesar, a system for code reviewing by a crowd of programmers.

Crowd computing raises new challenges at the intersection of computer systems and human-computer interaction, including minimizing latency, improving quality of work, and providing the right incentives to the crowd. The talk will discuss the design space and the techniques we have developed to address some of these problems. We are now in a position where “Wizard of Oz” is no longer just a prototyping technique — thanks to crowd computing, Wizard of Oz systems can be useful and deployable


Rob Miller is an associate professor of computer science at MIT, and associate director of the Computer Science and Artificial Intelligence Laboratory (CSAIL). He earned bachelors and masters degrees in computer science from MIT (1995) and PhD from Carnegie Mellon University (2002). He has won an ACM Distinguished Dissertation honorable mention, NSF CAREER award, and six best paper awards at UIST and USENIX. He has been program co-chair for UIST 2010, general chair for UIST 2012, and associate editor of ACM TOCHI. He has won two department awards for teaching, and was named a MacVicar Faculty Fellow for outstanding contributions to MIT undergraduate education. His research interests lie at the intersection of programming and human computer interaction: making programming easier for end-users (web end-user programming), making it more productive for professionals (HCI for software developers), and making people part of the programming system itself (crowd computing and human computation).

Random Sampling, Random Structures and Phase Transitions – Dana Randall

Dana Randall, Georgia Tech Wednesday, June 5, 2013 | Video


Sampling algorithms using Markov chains arise in many areas of computation, engineering, and science. The idea is to perform a random walk among the elements in a large state space so that samples chosen from the stationary distribution are useful for the application. In order to get reliable results efficiently, we require the chain to be rapidly mixing, or quickly converging to equilibrium. Often there is a parameter of the system (typically related to temperature or fugacity) so that at low values many natural chains converge rapidly while at high values they converge slowly, requiring exponential time. This dichotomy is often related to phase transitions in the underlying models. In this talk we will explain this phenomenon, giving examples from the natural and social sciences, including magnetization, lattice gasses, colloids, and models of segregation.


Dana Randall is the Advance Professor of Computing and an Adjunct Professor of Mathematics at the Georgia Institute of Technology. Her research in randomized algorithms focuses on the design and analysis of efficient algorithms for sampling and approximate counting, using techniques from computing, discrete mathematics and statistical physics. Dr. Randall received her A.B. in Mathematics from Harvard and her Ph.D. in Computer Science from U.C. Berkeley and held postdoctoral positions at the Institute for Advanced Study and Princeton. She is a Fellow of the American Mathematical Society, a National Associate of the National Academies, and the recipient of a Sloan Fellowship and an NSF Career Award.

Time Incentives in Public Procurement: Evidence from California and Minnesota – Greg Lewis

Greg Lewis, Harvard | Wednesday, May 22, 2013 | Video


Most procurement contracts incentivize timely delivery, either through the auction mechanism or the contract terms. We evaluate both of these approaches in the context of highway procurement, using data from California and Minnesota. We show that firms respond strongly to incentives: for example, in California, when contractors compete for contracts on the basis of both price and delivery date, contracts are completed 30-40% faster. We simulate counterfactual outcomes under different incentive schemes, and discuss the practical implications of our research for the design of procurement contracts.


Greg Lewis is associate professor of economics at Harvard University, and faculty research fellow at the National Bureau of Economic Research. His main research interests lie in industrial organization and market design, with a particular focus on auction theory and estimation. Recently, his time has been spent developing dynamic models of auction markets, suggesting methods for price discrimination in online display advertising, examining learning by firms in the British electricity market and analyzing how contracts terms interact with moral hazard in highway procurement. He received his bachelor’s degree in economics and statistics from the University of the Witwatersrand in South Africa, and his MA and PhD both from the University of Michigan.

Sum-Product Networks: Powerful Models with Tractable Inference – Pedro Domingos

Pedro Domingos, U Washington | Wednesday, May 8, 2013 | Video


Big data makes it possible in principle to learn very rich probabilistic models, but inference in them is prohibitively expensive. Since inference is typically a subroutine of learning, in practice learning such models is very hard. Sum-product networks (SPNs) are a new model class that squares this circle by providing maximum flexibility while guaranteeing tractability. In contrast to Bayesian networks and Markov random fields, SPNs can remain tractable even in the absence of conditional independence. SPNs are defined recursively: an SPN is either a univariate distribution, a product of SPNs over disjoint variables, or a weighted sum of SPNs over the same variables. It’s easy to show that the partition function, all marginals and all conditional MAP states of an SPN can be computed in time linear in its size. SPNs have most tractable distributions as special cases, including hierarchical mixture models, thin junction trees, and nonrecursive probabilistic context-free grammars. I will present generative and discriminative algorithms for learning SPN weights, and an algorithm for learning SPN structure. SPNs have achieved impressive results in a wide variety of domains, including object recognition, image completion, collaborative filtering, and click prediction. Our algorithms can easily learn SPNs with many layers of latent variables, making them arguably the most powerful type of deep learning to date. (Joint work with Rob Gens and Hoifung Poon.)


Pedro Domingos received an undergraduate degree (1988) and M.S. in Electrical Engineering and Computer Science (1992) from IST, in Lisbon. He received an M.S. (1994) and Ph.D. (1997) in Information and Computer Science from the University of California at Irvine. He spent two years as an assistant professor at IST, before joining the faculty of the University of Washington in 1999. He is the author or co-author of over 200 technical publications in machine learning, data mining, and other areas. He is a member of the editorial board of the Machine Learning journal, co-founder of the International Machine Learning Society, and past associate editor of JAIR. He was program co-chair of KDD-2003 and SRL-2009, and served on the program committees of AAAI, ICML, IJCAI, KDD, NIPS, SIGMOD, UAI, WWW, and others. He is a AAAI Fellow, and received a Sloan Fellowship, an NSF CAREER Award, a Fulbright Scholarship, an IBM Faculty Award, several best paper awards, and other distinctions.

Compressed Sensing and Natural Image Statistics – Yair Weiss

Yair Weiss, Hebrew U | Wednesday, April 24, 2013 | Video


Compressed sensing (CS) refers to a branch of applied mathematics which is based on the surprising result whereby signals that are exactly “k-sparse” (i.e. can be represented by at most k nonzero coefficients in some basis) can be exactly reconstructed using a small number of random measurements. Since natural images tend to be sparse in the wavelet basis, one of the motivating examples of CS has always been to reconstruct high resolution images from a small number of random measurements. Unfortunately, there are some significant deviations between the way that natural images behave and the assumptions of the dramatic theorems, and in fact random projections perform quite poorly when applied to real images. I will describe an alternative theory, which we call “Informative Sensing”, that seeks a small number of projections that are maximally informative given a known distribution over signals. I will show experimental results demonstrating that the informative projections indeed outperform random projections, but that the savings relative to more standard imaging methods are altogether rather modest.Joint work with Hyun Sung Chang and Bill Freeman.


Yair Weiss is a Professor of Computer Science and Engineering at the Hebrew University of Jerusalem. He is currently on sabbatical at Microsoft Research New England.

The Disruptive Power Of Three-Dimensional Printing – Deven Desai

Deven Desai, Thomas Jefferson School of Law | Thursday, May 2, 2013 | Video

*note the alternate date*


The Industrial Revolution was founded on economies of scale, but the next transformation in manufacturing may come from individual households. An additive (or 3D) printer is a desktop machine that can make customized physical objects from software and simple raw materials. This device promises to dramatically reduce the cost of making and distributing tangible goods, but it could also sharply increase patent infringement. Indeed, 3D printers present a challenge to patent law that is analogous to the disruption of copyright by MP3 files. This talk explores the implications of 3D printing for patents.


Deven Desai is a law professor at the Thomas Jefferson School of Law and recently completed serving as Academic Research Counsel at Google, Inc. As a law professor, he teaches trademark, intellectual property theory, business associations, and information privacy law. He is a graduate of the University of California, Berkeley and Yale Law School. He has also spent year as a Visiting Fellow at Princeton University’s Center for Information Technology Policy. Professor Desai’s scholarship examines how business interests and economic theories shape privacy and intellectual property law and where those arguments explain productivity or where they fail to capture society’s interest in the free flow of information and development. His articles include Speech Citizenry and the Market: A Corporate Public Figure Doctrine 98 Minnesota Law Review __ (2013) (forthcoming); Bounded by Brands: An Information Network Approach to Brands, U.C. Davis Law Review (2013) (forthcoming); Beyond Location: Data Security in the 21st Century, Communications of the ACM (January, 2013); Response: An Information Approach to Trademarks, 100 Georgetown Law Journal 2119 (2012); From Trademarks to Brands, 46 Florida Law Review 981 (2012); The Life and Death of Copyright, 2011 Wisconsin Law Review 219 (2011); Brands, Competition, and the Law, 2010 Brigham Young Law Review 1425 (2010) (with Spencer Waller); Privacy? Property?: Reflections on the Implications of a Post-Human World 18 Kansas J. of Law & Public Policy (2009); Property, Persona, and Preservation, 81 Temple Law Review 67 (2008); and Confronting the Genericism Conundrum, 28 Cardozo Law Review 789 (2007) (Sandra L. Rierson, co-author).

How Entrepreneurs Came to Own Innovation: The Rhetoric of Economic Risk in High-Tech – Gina Neff

Gina Neff, U of Washington/Princeton | Wednesday, April 10, 2013 | Video


How did innovation come to be synonymous with entrepreneurship? How did creativity become equated with risk? Perhaps more importantly, how did these concepts lead to advice such as that given by New York Times columnist Thomas Friedman: “Need a Job? Then Invent One?”This talk will present research on the first wave of employees with dot-com start-ups of the 1990s and 2000s who exhibited entrepreneurial behavior in their jobs–investing time, energy, and other personal resources–when they themselves were employees and not entrepreneurs. I argue that this “venture labor” is part of a longer and broader social shifting of economic risk to individual responsibility and understanding it is of paramount importance for encouraging innovation and, even more important, for creating sustainable work environments in high-tech sectors today.


Gina Neff is an associate professor of communication at the University of Washington. She studies the contemporary economics of media production and the impact of new technologies on communication, focusing on both high-tech and media industries. Her book Venture Labor: Work and the Burden of Risk in Innovative Industries (MIT 2012) examines the risk and uncertainties borne by New York City’s new media pioneers during the first dot-com boom. She co-directs the Project on Communication Technology and Organizational Practices, a research group studying the roles of communication technology in the work around building design and construction. Her research has been funded by the National Science Foundation, and she is currently at work on a three-year project funded by Intel studying the impact of social media and consumer health technologies on the organization of primary care.

She holds a Ph.D. in sociology from Columbia University, where she remains an external faculty affiliate of the Center on Organizational Innovation. She is currently a fellow at Princeton’s Center for Information Technology Policy and visiting scholar at NYU’s Media, Culture and Communication department. She has held appointments at UC San Diego, UCLA, and Stanford University. In addition to academic outlets, her research and writing have been featured in The New York Times, Christian Science Monitor, Fortune, The American Prospect, and The Nation.

So You Think Quantum Computing Is Bunk? – Scott Aaronson

Scott Aaronson, MIT | Wednesday, April 10, 2013 | Video


In this talk, I’ll take an unusual tack in explaining quantum computing to a broad audience. I’ll start by assuming, for the sake of argument, that scalable quantum computing is “too crazy to work”: i.e., that it must be impossible for some fundamental physical reason. I’ll then investigate the sorts of radical additions or changes to current physics that we seem forced to contemplate in order to justify such an assumption. I’ll point out the many cases where such changes seem ruled out by existing experiments, or by no-go theorems such as the Bell Inequality. I’ll also mention two recent no-go theorems for so-called “epistemic” hidden-variable theories: one due to Pusey, Barrett, and Rudolph, the other to Bouland, Chua, Lowther, and myself. Finally, I’ll discuss my 2004 notion of a “Sure/Shor separator,” as well as the BosonSampling proposal [A.-Arkhipov 2011] and its recent experimental realizations—which suggest one possible route to falsifying the Extended Church-Turing Thesis more directly than by building a universal quantum computer.


Scott Aaronson is the TIBCO Career Development Associate Professor of Electrical Engineering and Computer Science at MIT. His research focuses on the capabilities and limits of quantum computers, and computational complexity theory more generally. His book, “Quantum Computing Since Democritus,” was recently published by Cambridge University Press; he’s also written about quantum computing for Scientific American and the New York Times. He’s received the National Science Foundation’s Alan T. Waterman Award, as well as MIT’s Junior Bose Award for Excellence in Teaching.

Platforms, practices, politics: Towards an open history of social media – Jean Burgess

Jean Burgess, Queensland U of Technology | Wednesday, March 27, 2013 | Video


Social media has been with us as a mainstream phenomenon for barely a decade now. That period has seen multiple, distinct paradigm shifts in the business models, uses, and discourses surrounding social media, as well as in approaches to conducting research on and through particular social media platforms. In this paper I draw on recent attempts within media, communication and cultural studies to go beyond static, single-platform snapshots and to develop more synthesized, general accounts of how social media has evolved since the early 2000s. I show how we might identify patterns of change across platforms and over time, and discuss the practical and conceptual challenges of opening up these short but dynamic histories of the proprietary web.


Jean Burgess is an Associate Professor of Digital Media Studies and Deputy Director of the ARC Centre of Excellence for Creative Industries & Innovation (CCI) at Queensland University of Technology, Australia. Her research focuses on the uses, politics and methodological implications of social and mobile media platforms.

The Virtual Lab – Duncan Watts

Duncan Watts, Microsoft Research New York CityWednesday, December 5, 2012 | Video


Crowdsourcing sites like Amazon’s Mechanical Turk are increasingly being used by researchers to construct “virtual labs” in which they can conduct behavioral experiments. In this talk, I describe some recent experiments that showcase the advantages of virtual over traditional physical labs, as well as some of the limitations. I then discuss how this relatively new experimental capability may unfold in the near future, along with some implications for social and behavioral science.


Duncan Watts is a principal researcher at Microsoft Research and a founding member of the MSR-NYC lab. From 2000-2007, he was a professor of Sociology at Columbia University, and then prior to joining Microsoft, a principal research scientist at Yahoo! Research, where he directed the Human Social Dynamics group . He has also served on the external faculty of the Santa Fe Institute and is currently a visiting fellow at Columbia University and at Nuffield College, Oxford.His research on social networks and collective dynamics has appeared in a wide range of journals, from Nature, Science, and Physical Review Letters to the American Journal of Sociology and the Harvard Business Review. He is also the author of three books, including Six Degrees: The Science of a Connected Age (W.W. Norton, 2003) and Everything is Obvious*: Once You Know The Answer (Crown Business, 2011).

He holds a B.Sc. in Physics from the Australian Defence Force Academy, from which he also received his officer’s commission in the Royal Australian Navy, and a Ph.D. in Theoretical and Applied Mechanics from Cornell University.

The Applicant Auction for Top-Level Domains: Using an auction to efficiently resolve conflicts among applicants – Peter Cramton

Peter Cramton, University of Maryland | Wednesday, November 28, 2012 | Video


The prospect of using auctions to resolve conflicts among parties competing for the same top-level internet domains is described. In such an auction the winner’s payment is divided among the losers, whereas if the conflict is not resolved then ICANN will conduct an auction and retain the winner’s payment. For first-price and second-price sealed-bid auctions, we characterize equilibrium bidding strategies and provide examples, assuming bidders’ valuations are distributed independently and are either symmetrically or asymmetrically distributed. The qualitative properties of equilibria reveal novel features; for example, in a second-price auction a bidder might bid more than her valuation in order to drive up the winner’s payment. Even so, examples indicate that in symmetric cases a bidder’s expected profit is the same in the two auction formats. We then test in the experimental lab two auction formats that extent the setting from a single domain to the actual setting with many domains. The first format is a sequential first-price sealed-bid auction; the second format is a simultaneous ascending clock auction. The framing and subjects were chosen to closely match the actual setting. Subjects were PhD students at the University of Maryland in Economics, Computer Science, and Computer Engineering, with training in game theory and auction theory. Each subject played the role of an actual company (e.g., Google) and bid for domains (e.g., .book) consistent with the company’s applications. Subjects were given instructions explaining the auction and the equilibrium theory for the single-item case in relevant examples. Both formats achieved auction efficiencies of 98% in the lab. This high level of efficiency is especially remarkable in the case with asymmetric distributions—the format performed better than the simple single-item equilibrium despite the presence of budget constraints in the lab. This experiment together with previous results on the robustness of ascending auctions in general and simultaneous ascending clock auctions in particular suggest that the simultaneous ascending clock auction will perform best in this setting.

See “Applicant Auctions for Internet Top-Level Domains: Resolving Conflicts Efficiently” (with Ulrich Gall, Pacharasut Sujarittanonta, and Robert Wilson), Working Paper, University of Maryland, 11 November 2012. [Presentation]


Peter Cramton is Professor of Economics at the University of Maryland. Since 1983, he has conducted research on auction theory and practice. This research appears in the leading economics journals. The main focus is the design of auctions for many related items. Applications include spectrum auctions, electricity auctions, and treasury auctions. On the practical side, he is Chairman of Market Design Inc., an economics consultancy founded in 1995, focusing on the design of auction markets. He also is Founder and Chairman of Cramton Associates LLC, which since 1993 has provided expert advice on auctions and market design. Since 2001, he has played a lead role in the design and implementation of electricity auctions in France and Belgium, gas auctions in Germany, and the world’s first auction for greenhouse gas emissions held in the UK in 2002. He has advised numerous governments on market design and has advised dozens of bidders in high-stake auction markets. Since 1997, he has advised ISO New England on electricity market design and was a lead designer of New England’s forward capacity auction. He led the design of electricity and gas markets in Colombia, including the Firm Energy Market, the Forward Energy Market, and the Long-term Gas Market. Since June 2006, he played a leading role in the design and development of Ofcom’s spectrum auctions in the UK. He has advised the UK, the US, and Australia on greenhouse gas auction design. He led the development of the FAA’s airport slot auctions for the New York City airports. He received his B.S. in Engineering from Cornell University and his Ph.D. in Business from Stanford University.

Miku: Virtual Idol as Media Platform – Ian Condry

Ian Condry, MIT | Wednesday, November 14, 2012 | Video


Miku Hatsune is Japan’s number one virtual idol. Her songs are sold online, she is one of the most requested karaoke downloads, she promotes Toyota in TV commercials, she performs concerts with live bands — and she doesn’t exist. Miku is a voice in music synthesizer software, and her community of users have created something new in the world of popular culture: a crowd-sourced celebrity. Based on fieldwork in Japan and the US, this talk will explore the dynamics of the social in media and the value of collaborative creativity.


Ian Condry is a cultural anthropologist and associate professor of Comparative Media Studies at MIT. His forthcoming book The Soul of Anime: Collaborative Creativity and Japan’s Media Success Story (January 2013, Duke University Press) focuses on Japan’s anime creators including participant-observation in studios, fan conventions and toy companies. His first book, Hip-Hop Japan: Rap and the Paths of Cultural Globalization (2006) is based on fieldwork in Tokyo nightclubs and recording studios. More info:

Music intelligence & the 'Taste Profile' - What computers think of you and your music taste – Brian Whitman

Brian Whitman, The Echo Nest | Wednesday, October 31, 2012 | Video


Over 200 million people now trust an algorithm they’ve never met to listen to and discover music. But music needs a bit more care than collaborative filtering or automated editorial approaches can give, and before we let Facebook automatically make mixtapes for our crushes, we should step back and see what the potential of music analysis is and how we can give it more respect.For the past 10 years I’ve been working on automatic music analysis, first academically and now as the co-founder and CTO of the Echo Nest, a company you’ve never heard of but powers most music discovery experiences you have on the internet today, from Spotify to Clear Channel to MTV. I’ll show how the interaction between listeners and music is being modeled today, where it is amazing and where it falls flat, and how connections are being made between your music taste and your identity.


Brian is recognized as a leading scientist in the area of music and text retrieval and natural language processing. He received his doctorate from MIT’s Media Lab in 2005 and co-founded The Echo Nest to provide music recommendation, search, playlisting, fingerprinting and personalization technology based on his research to much of the online music industry. As the CTO of the Echo Nest, Brian leads new product development and focuses on future taste profile and music analytic products.

Real applications of non-real numbers – Roger Myerson

Roger Myerson, The University of Chicago | Thursday, October 18, 2012*

*Special date & time*


This paper considers a simple model of credit cycles driven by moral hazard in financial intermediation. Investment advisors or bankers must earn moral-hazard rents, but the cost of these rents can be efficiently spread over a banker’s entire career, by promising large back-loaded rewards if the banker achieves a record of consistently successful investments. The dynamic interactions among different generations of bankers can create equilibrium credit cycles with repeated booms and recessions. We find conditions when taxing workers to subsidize bankers can increase investment and employment enough to make the workers better off.

The paper is at


Roger Myerson is the Glen A. Lloyd Distinguished Service Professor of Economics at the University of Chicago. He has made seminal contributions to the fields of economics and political science. In game theory, he introduced refinements of Nash’s equilibrium concept, and he developed techniques to characterize the effects of communication when individuals have different information. His analysis of incentive constraints in economic communication introduced some of the fundamental ideas in mechanism design theory, including the revelation principle and the revenue-equivalence theorem in auctions and bargaining. Professor Myerson has also applied game-theoretic tools to political science, analyzing how political incentives can be affected by different electoral systems and constitutional structures.

Myerson is the author of Game Theory: Analysis of Conflict (1991) and Probability Models for Economic Decisions (2005). He also has published numerous articles in Econometrica, the Journal of Economic Theory, Games and Decisions, and the International Journal of Game Theory, for which he served as an editorial board member for 10 years.

Professor Myerson has a PhD from Harvard University and taught for 25 years in the Kellogg School of Management at Northwestern University before coming to the University of Chicago in 2001. He is a member of the American Academy of Arts and Sciences and of the National Academy of Sciences. In 2007, he was awarded the 2007 Nobel Memorial Prize in Economic Sciences in recognition of his contributions to mechanism design theory.

Real applications of non-real numbers – Alex Lubotzky

Alex Lubotzky, Hebrew U of Jerusalem | Wednesday, October 10. 2012 | Video


The system of real numbers are defined mathematically as a “completion’ of the rational numbers. But this is not the only way to do it! In fact there are infinitely many others completions- the so called “p-adic numbers”. These numbers were defined for pure mathematical reasons and have been a subject of research for a century. But in the last 3 decades they have found ‘real world’ applications in computer science, construction of networks, algorithms etc. We will try to tell the story in a way which hopefully will make sense also to non-mathematicians.


Alex Lubotzky is the Weil Professor of Mathematics at the Hebrew University of Jerusalem and an adjunct prof. of math at Yale University. He got his PhD. from Bar-Ilan University in 1980. Following an army service he joined the Hebrew University in 1983.His main area of research is group theory which he likes to combine with other areas like geometry, number theory, combinatorics and computer science. One of his best known works is the construction of Ramanujan graphs (which are optimal expanders) jointly with Phillips and Sarnak. This opened a world of connections between graph theory and representation theory. Lubotzky is an Honorary Foreign Member of the American Academy of Arts and Science and in 2006 he received an honorary degree from the University of Chicago for his contributions to modern mathematics.

Diffusion of Microfinance – Matt Jackson

Matt Jackson, Stanford | Wednesday, October 3. 2012 | Video


We examine how participation in a microfinance program diffuses through social networks, using detailed demographic, social network, and participation data from 43 villages in South India. We exploit exogenous variation in the importance (in a network sense) of the people who were first informed about the program, the “injection points” .Microfinance participation is significantly higher when the injection points have higher eigenvector centrality. We also estimate structural models of diffusion that allow us to (i) determine the relative roles of basic information transmission versus other forms of peer influence, and (ii) distinguish information passing by participants and nonparticipants. We find that participants are significantly more likely to pass informationon to friends and acquaintances than informed non-participants. However, information passing by non-participants is still substantial and significant, accounting for roughly one-third of informedness and participation. We also find that, once we have properly conditioned on an individual being informed, her decision to participate is not significantly affected by the participation of her acquaintances.


Matthew O. Jackson is the Eberle Professor of Economics at Stanford University and an external faculty member of the Santa Fe Institute and a fellow of CIFAR. Jackson’s research interests include game theory, microeconomic theory, and the study of social and economic networks, including diffusion, learning, and network formation. He was at Northwestern and Caltech before joining Stanford, and has a PhD from Stanford and BA from Princeton. Jackson is a Fellow of the Econometric Society and the American Academy of Arts and Sciences, and former Guggenheim Fellow.

Duolingo: Learn a Language for Free While Helping to Translate the Web – Luis von Ahn

Luis von Ahn, Carnegie Mellon University | Wednesday, September 19. 2012 | Video


I want to translate the Web into every major language: every webpage, every video, and, yes, even Justin Bieber’s tweets. With its content split up into hundreds of languages — and with over 50% of it in English — most of the Web is inaccessible to most people in the world. This problem is pressing, now more than ever, with millions of people from China, Russia, Latin America and other quickly developing regions entering the Web. In this talk, I introduce my new project, called Duolingo, which aims at breaking this language barrier, and thus making the Web truly “world wide.”

We have all seen how systems such as Google Translate are improving every day at translating the gist of things written in other languages. Unfortunately, they are not yet accurate enough for my purpose: Even when what they spit out is intelligible, it’s so badly written that I can’t read more than a few lines before getting a headache.

With Duolingo, our goal is to encourage people, like you and me, to translate the Web into their native languages.


Luis von Ahn is the A. Nico Habermann Associate Professor of Computer Science at Carnegie Mellon University. He is working to develop a new area of computer science that he calls Human Computation, which aims to build systems that combine the intelligence of humans and computers to solve large-scale problems that neither can solve alone. An example of his work is reCAPTCHA, in which over one billion people — 15% of humanity — have helped digitize books and newspapers. Among his many honors are a MacArthur Fellowship, a Packard Fellowship, a Sloan Research Fellowship, a Microsoft New Faculty Fellowship, the ACM Grace Hopper Award, and CMU’s Herbert A. Simon Award for Teaching Excellence and Alan J. Perlis Teaching Award. He has been named one of the “50 Best Brains in Science” by Discover Magazine, one of the 50 most influential people in technology by, and one of the “Brilliant 10 Scientists” by Popular Science Magazine.

How users evaluate things and each other in social media – Jure Leskovec

Jure Leskovec, Stanford University | Wednesday, September 5. 2012 | Video


In a variety of domains, mechanisms for evaluation allow one user to say whether he or she trusts another user, or likes the content they produced, or wants to confer special levels of authority or responsibility on them. We investigate a number of fundamental ways in which user and item characteristics affect the evaluations in online settings. For example, evaluations are not unidimensional but include multiple aspects that all together contribute to user’s overall rating. We investigate methods for modeling attitudes and attributes from online reviews that help us better understand user’s individual preferences. We also examine how to create a composite description of evaluations that accurately reflects some type of cumulative opinion of a community. Natural applications of these investigations include predicting the evaluation outcomes based on user characteristics and to estimate the chance of a favorable overall evaluation from a group knowing only the attributes of the group’s members, but not their expressed opinions


Jure Leskovec is assistant professor of Computer Science at Stanford University where he is a member of the Info Lab and the AI Lab. His research focuses on mining large social and information networks.Problems he investigates are motivated by large scale data, the Web and on-line media. This research has won several awards including best paper awards at KDD (2005, 2007, 2010), WSDM (2011), ICDM (2011) and ASCE J. of Water Resources Planning and Management (2009), ACM KDD dissertation award (2009), Microsoft Research Faculty Fellowship (2011), Alfred P. Sloan Fellowship (2012) and NSF Early Career Development (CAREER) Award (2011). He received his bachelor’s degree in computer science from University of Ljubljana, Slovenia, Ph.D. in machine learning from the Carnegie Mellon University and postdoctoral training from Cornell University. You can follow him on Twitter @jure

Playing 'Hide and Seek' - The hidden genome – Michal Linial

Michal Linial, The Hebrew University of Jerusalem, Israel | Wednesday, August 30. 2012 | Video


The overwhelming increase in sequencing methodology resulted in the accumulation of millions of DNA sequences. These sequences are collected from thousands of genomes that (ideally) sample the ‘tree of life’. I will briefly discuss the ‘minimal set of instructions’ by which a linear sequence is transformed into a functional protein. What happen when the statistical noise is too high, thus classical procedures to predict protein sequences fail? I will focus on the challenge of identifying short proteins that remain buried in the genomic data. For illustration, I will take you for a ‘treasure hunt’ for short proteins.

Many short proteins share fuzzy features that are common to most animal venom. I will discuss the limitation in using classical tools that are based on string comparison, or pattern finding to identify short proteins. For this task, statistical machine learning methods were useful in identifying hidden bioactive sequences in several genomes. Evidently, such sequences are attractive candidates for novel therapy. The test case of short proteins illustrates the importance of a cycle that starts by a biological hypothesis, then uses a computational formulation and finalizes by an experimental validation. Finally, I will discuss our genomes with respect to our ‘partners’ (viruses, bacteria). Once the interaction of these genomes is considered, the source for the dynamic nature of human evolution becomes evident. Related publications:

  • Rappoport N, Karsenty S, Stern A, Linial N, Linial M. (2012) Nucl. Acids Res. 40:D313-D320
  • Rappoport N, Linial M. (2012) PLoS Comput Biol. 8:e1002364.
  • Naamati G, Askenazi M, Linial M. (2010) Bioinformatics 26:i482-i488.
  • Naamati G, Askenazi M, Linial M (2009) Nucl. Acids Res. 37:W363-368.
  • Kaplan N, Morpurgo N, Linial M. (2007) J Mol Biol. 369:553-566.


Michal Linial is a Professor of Biochemistry, The Hebrew University, Jerusalem, Israel and a Director of the SCCB, the Sudarsky Center for Computational Biology.ML had published over 150 scientific papers and abstracts on diverse topics in molecular biology, cellular biology, bioinformatics, neuroscience the integration of tools to improve knowledge extractions. M. Linial has an experimental and computational laboratory. M.L is the leader and the founder of the first established educational program in Israel for Computer Science and Life Science (from 1999) for Undergraduate-Graduate studies. Her expertise in the synapse let to the study of protein families, protein-protein interactions with a global view on protein networks and their regulation. Molecular biology, cell biology and biochemical methods are applied in all research initiated in her laboratory. She and her laboratory are developing new computational and technological tools for large-scale cell biological research M. Linial and her colleagues apply MS based and genomics (DNA Chip) approaches for studying changes in neuronal development, and disease oriented research. She published over 180 scientific papers including book chapters and numerous reviews.The solid informatics approaches are used for large database storage and constant updating of several systems in view of classification, validation and functional predictions. M.L. and her students has been an active participant in NIH structural genomics initiatives and she participated in Structural Genomics effort Task for target selections. She and her colleagues have created several global classification systems that are used by the biomedical and biology communities. Most notably are the ProtoNet, EVEREST, ProTarget and PANDORA, mirror, ClanTox and more. All those developed web systems are provided as an open source for investigators.

Dynamic Games with Asymmetric Information: A Framework for Empirical Work – Ariel Pakes

Ariel Pakes, Harvard University | Wednesday, August 29. 2012 | Video


We develop a framework for the analysis of dynamic games that can be applied to the analysis of firm which compete in a market whose characteristics evolve over time as a probabilistic function of the actions of the firms competing in that market. Firm’s chose their actions to maximize their perceptions of the discounted value of the returns that will accrue to them as a result of those actions. These returns depend on both their own states and their competitor’s states. The firms know their own states, but only observe imprecise signals on the states of their competitors. Our goal is to provide a framework capable of analyzing the impact of policy or environmental changes in such a setting. Bayesian perfect Nash equilibria for environments that are rich enough to adequately approximate behavior have computational and informational demands that both; (i) make them impossible for applied researchers to use, and (ii) unlikely to be the best approximation to agent’s actual behavior. So we introduce an alternative notion of equilibria which is less demanding of both agents and researchers, while still implying agents “optimize” in a meaningful sense of that word. We show that: (i) there is an artificial intelligence algorithm that makes it relatively easy to compute (at least some of) the resultant equilibria, and (ii) it is relatively easy to use the properties of that equilibria to estimate any unknown parameters of the game. We use the analysis of a de-regulated electric utility market as an example. Two firms each own several generators and bid “supply functions” into the market in every period (a quantity supplied as an increasing function of price). An independent system operator (an ISO) sums the supply curves horizontally and intersects the result with demand to determine the period’s price and the quantities to be produced by each firm. The firm’s cost of supplying electricity on each of its generators is increasing in the current quantity produced and stochastically increasing in the quantities produced since the last time the firm did maintenance on that generator. Firms do not know the current cost of their competitor’s generators but realize that the returns they will earn from the bid on each of its generator will increase the less the quantity supplied by other generators (their own, as well as those of its competitors). This provides incentives for firms to simply shut down some generators without doing maintenance, and to implicitly co-ordinate shutdowns across firms. Consumers pay the price through the resultant increase in the price of electricity. Joint work with Chaim Fershtman


Ariel Pakes is the Steven McArthur Heller Professor of Economics in the Department of Economics at Harvard University, where he teaches courses in Industrial Organization and in Econometrics. Before coming to Harvard in 1999, he was the Charles and Dorothea Dilley Professor of Economics at Yale University (1997-99). He has held other tenured positions at Yale (1988-97), the University of Wisconsin (1986-88), and the University of Jerusalem (1985-86). Pakes received his doctorate degree from Harvard University in 1980, and he stayed at Harvard as a Lecturer until he took up a position in Jerusalem in 1981. Pakes received the award for the best graduate student advisor at Yale in 1996. Pakes was elected fellow of the American Academy of Arts and Sciences in 2002. He received the Frisch Medal of the Econometric Society in 1986, was elected as a fellow of that society in 1988, and gave the Fisher-Schultz lecture at the World Congress of that society in 2005. He was the Distinguished Fellow of the Industrial Organization Society in 2007. He has been on the editorial boards of the RAND Journal of Economics, Econometrica, Economic Letters and the Journal of Economic Dynamics and Control. He is also a research associate of the NBER, and has been member of the AEA Committee on Government Statistics, the chair of the AEA Census Advisory Panel, and co-editor of a Proceedings of the National Academy of Science issue on “Science, Technology and the Economy”. Professor Pakes’ research has been in Industrial Organization (I.O.), the Economics of Technological Change and in Econometric Theory. He and his co-authors have focused on developing techniques which allow us to analyze market responses to policy and environmental changes. This includes; econometric work on how to estimate demand and cost systems and then use the estimated parameters to analyze equilibrium responses in different institutional settings, empirical work which uses these techniques to analyze market outcomes in different industries, and theoretical work developing frameworks for the applied analysis of dynamic oligopolies (with and without collusive possibilities, and with and without asymmetric information).

Gender, Competitiveness and Career Choices – Muriel Niederle

Muriel Niederle, Stanford University | Wednesday, August 22. 2012 | Video


Gender differences in competitiveness are often discussed as potential explanation for gender differences in labor market outcomes. We correlate an incentivized measure of competitiveness with the first important career choice of secondary school students in the Netherlands. At the age of 15, these students have to pick one out of four study profiles, which vary in how prestigious they are. While boys and girls have very similar levels of academic ability, boys are substantially more likely than girls to choose more prestigious profiles. We find that 25% of this gender difference can be attributed to gender differences in competitiveness. This lends support to the extrapolation of laboratory findings on competitiveness to labor market settings. Joint work with Hessel Oosterbeek and Thomas Buser.


Muriel is a Professor of Economics at Stanford. In her own words: I am an experimental economist, and as such, have some experiments that fall outside my main areas of gender or market design. Most recently, I got interested in k-level models. The first strand of literature I am working on can be broadly thought of as market design. While that includes studying markets that have been designed (such as the National Residency Matching Market), I am also interested redesigning markets, or adding features such as signaling to help markets such as the economics job market work better. Most recently, I have been getting involved in working with the San Francisco Unified School District to help redesign their school choice system. In market design, I have used theory, experiments, as well as data collected by others. My second strand of work is work on gender differences. So far, I have only experimental papers in that are, showing that women may not be as competitive as men, especially when they have to compete against men.

Wireless Spectrum Sharing: Opportunities for Interdisciplinary Research – Anant Sahai

Anant Sahai, University of California, Berkeley | Wednesday, August 8, 2012 | Video


Under the current static system of frequency assignment, a great deal of spectrum remains underused. This seeming waste represents an opportunity for frequency-agile cognitive radios to improve performance. Understanding this opportunity forces us to take a closer look at the whole question of “regulatory overhead.” Until recently, cognitive radios represented the “Medical Marijuana” of wireless research — rhetoric on both sides characterized by distrust, wishful thinking, and vested interests, but the underlying “technology” in question was still very much illegal. Regulatory changes were required before research in this area could truly impact practice. Recent steps taken by the FCC in the TV Whitespaces demonstrate that the government is serious about change, and just last month, the President’s Council of Advisers on Science and Technology (PCAST) released a report that advocated expanding this approach beyond the TV bands. However, the problem is that while we have a rough sense of what we want to achieve at a high level, as a community, we do not yet know what this regulatory change should entail at the detailed level and more troubling, even how we would recognize the right answer if we saw it.The full scope of the problem weaves together information theory, signal processing, economics, and law in a nontrivial way (and probably also cryptography and social networks). In this talk, I will give an introduction to the opportunity in the context of the TV Whitespaces. I’ll use some simulations based on real FCC data and realistic propagation models to give a quantitative sense of the tradeoffs involved, and then show idealized models that enable a conceptual understanding of the “overhead” in the context of spectrum sensing. I will then elucidate what “light handed regulation” could mean in the cognitive radio context, giving a simple criminal-law inspired model to reveal something about the overhead and tradeoffs involved. I’ll close with some interesting future research directions.


Anant Sahai (BS ’94 UC Berkeley, MS ’96 MIT, PhD ’01 MIT) is an Associate Professor in the Department of Electrical Engineering and Computer Sciences at the University of California at Berkeley, where he joined the faculty in 2002. He is a member of the Berkeley Wireless Research Center (BWRC) and the Wireless Foundations Center (WiFo). In 2001, he spent a year at the wireless startup Enuvis developing adaptive signal processing algorithms for extremely sensitive GPS receivers implemented using software defined radio. Prior to that, he was a graduate student at the Laboratory for Information and Decision Systems (LIDS) at the Massachusetts Institute of Technology (MIT). His research interests are in wireless communication, decentralized control, and information theory. He is particularly interested in delay, feedback, and complexity from an information-theoretic perspective and in cognitive radio from a regulatory perspective.

The Wonders of the Probabilistic Method – Nati Linial

Nati Linial, Hebrew University | Wednesday, August 8, 2012 | Video


will try to explain some key principles in modern mathematics which combine ideas from combinatorics and probability. In particular I will emphasize the surprising role that probability theory plays in the study of combinatorics. How it allows us to investigate complicated graphs and networks without having to reveal all the specific details of individual large graphs or networks. This talk is intended for a general audience. The necessary mathematical background is at the level of good high-school education.


Nati Linial is a professor of computer science at the Hebrew University of Jerusalem. In his own words: I got my undergraduate education in mathematics at the Technion. I did my PhD at the Hebrew University with a thesis in graph theory. Following a postdoctoral period at UCLA math I joined the faculty of the Hebrew University. My main areas of interest are combinatorics, theoretical computer science and bioinformatics. I had about 30 graduate students so far (currently I have 7 PhD students and one MSc student) I am married to Michal, a life-science professor at the Hebrew University. We have three children who are, respectively, an artist, a poet and a budding physicist. I like long-distance running, reading and classical music.

Lines, Shading, and the Perception of 3D Shape – Ted Adelson

Ted Adelson, MIT | Wednesday, August 1, 2012 | Video


Humans can easily see 3D shape from single 2D images, exploiting multiple kinds of information. There are several subfields (in both human vision and computer vision) devoted to the study of particular cues to 3D shape, such as shading, texture, and contours. However, the resulting algorithms remain specialized and fragile (in contrast with the flexibility and robustness of human vision). Recent work in graphics and psychophysics has demonstrated the importance of local orientation structure in conveying 3D shape. This information is fairly stable and reliable across rendering condition. We have developed an exemplar-based system (which we call Shape Collage) that learns to associate image patches with corresponding 3D shape patches.

We train it with synthetic images of “blobby” objects rendered in various ways, including solid texture, Phong shading, and line drawings. Given a new image, it finds the best candidate scene patches and assembles them into a coherent interpretation of the object shape. Our system is the first that can retrieve the shape of naturalistic objects from line drawings. The same system, without modification, works for shape-from-texture and can also do shape-from-shading without requiring Lambertian surfaces. Thus disparate types of image information can be processed by a single mechanism to extract 3D shape.

(Collaborative work with Forrester Cole, Phillip Isola, Fredo Durand, and William Freeman.)

Random Graph Models of Kidney Exchange – Al Roth

Al Roth, Harvard University | Wednesday, July 25. 2012 | Video


Kidney exchange involves creating a non-monetary marketplace through which incompatible donors and patients can take part in exchanges, so that each patient in the exchange receives a transplant from a compatible donor. I’ll recount the brief history of kidney exchange, explain some of the technical problems that have been overcome, and some which remain.

Barack Obama and the politics of social media for national policy-making – James Katz

James Katz, Boston University | Wednesday, October 15, 2011


Social media help people do most everything, ranging from meeting new friends and finding new restaurants to overthrowing dictatorships. This includes political campaigning; one need look no further than Barack Obama’s successful presidential campaigns to see how these communication technologies can alter the way politics is conducted. Yet social media have not had much import for setting national policy as part of regular administrative routines. This is the case despite the fact that, since his election in 2008, President Obama has on several occasions proclaimed that he wanted his administration to draw on social media to make the federal government run better. While there have been some modifications to governmental procedures due to the introduction of social media, the Obama administration practices have fallen far short of its leader’s audacious vision. Despite voluminous attention to social media in other spheres of activity, there has been little to point to in terms of successfully drawing on the public to help set national policies. What might account for this? I try to answer this question in my talk by exploring the attempts by the Obama White House to use social media tools and the consequences arising from such attempts. I also suggest some potential reasons behind the particular uses and outcomes that have emerged in terms of presidential-level social media outreach. As part of my conclusion, I outline possible future directions.


James E. Katz, Ph.D., is the Feld Family Professor of Emerging Media at Boston University’s College of Communication where he directs its Center for Mobile Communication Studies and Division of Emerging Media. His research on the internet, social media and mobile communication has been internationally recognized, and he is frequently invited to address high-level industry, governmental and academic groups on his research findings. His latest book, with Barris and Jain, is The Social Media President: Barack Obama and the Politics of Citizen Engagement on which this talk is based.

Arrival Guidance

Upon arrival to One Memorial Drive, visitors should be prepared to show a valid photo ID and sign the Building Visitor Log when approaching the Lobby Floor Security Desk. Alert them to the name of the event you are attending and ask them to direct you to the appropriate floor. Typically, the talks are located in the Conference Center on the first floor, however sometimes the location may change.

*Hospitality Notice: Microsoft Research may provide hospitality at this event. Because different universities and legal jurisdictions have differing rules, we rely on you to know whether acceptance of this invitation would be inconsistent with those rules. Accordingly, by accepting our invitation, you confirm that this invitation is compliant with your institution’s policies.

If you have any questions or concerns, please send us an email.


Past Speaker Series at the MSR Lab

For past Microsoft Research colloquium series presentations, please visit: