Privacy-Enhanced Personalization


January 16, 2006


Alfred Kobsa


University of California, Irvine


Web personalization has demonstrated to be advantageous for both online customers and vendors. However, its benefits are counteracted by privacy concerns. Personalized systems need to take these concerns into account, as well as privacy laws and industry self-regulations that may be in effect. We present two research endeavors in this framework of “privacy-enhanced personalization”: Effective disclosure of privacy practices. Current “privacy policies” are ineffective in allaying privacy concerns. They are written in a lengthy and legalistic manner, and in effect, hardly ever read by Internet shoppers. We tested a different, HCI-oriented approach, namely web design templates in which every entry field for personal data is accompanied by a clear and concise explanation of how the retailer will deal with the respective piece of data, and what benefits customers can expect from sharing this personal information. We compared users of an online book retail website that used a traditional privacy disclosure with users of the same website after it was redesigned based on our templates. Subjects in the second group not only rated its privacy practices significantly higher, answered 8% more questions and gave 20% more answers, but also rated the perceived benefit resulting from data disclosure significantly higher and bought books 33% more often. Adjusting personalization privacy constraints. Privacy concerns, privacy laws and self-regulation not only affect the personal data that can be collected, but also frequently-employed personalization methods. Our research aims at maximizing the personalization benefits, while at the same time satisfying the privacy constraints that prevail during a user session. We exploit the ability of software product lines to support software variability and developed a user modeling architecture that supports architectural level configuration management to dynamically select those personalization methods that satisfy the current privacy constraints. We describe a pilot experiment with an existing user modeling server and a software-architecture based development environment.


Alfred Kobsa

Dr Alfred Kobsa is a Professor in the Donald Bren School of Information and Computer Sciences of the University of California, Irvine. Before he was a Director of the Institute for Applied Information Technology (FIT) at the German National Research Center for Information Technology (GMD), and a Professor of Computer Science at the University of Essen, Germany. He was also an Associate Professor of Information Systems at the Department of Information Science at the University of Konstanz, Germany, and a Senior Researcher at the Department of Computer Science of the University of Saarbr├╝cken. He received his master degrees in Computer Science and in the Social and Economic Sciences from the Johannes Kepler University Linz, Austria, and his Ph.D. in Computer Science from the University of Vienna, Austria and the Vienna University of Technology. Dr Kobsa’s research lies in the areas of user modeling and personalized systems (with applications in the areas of information environments, expert finders, and user interfaces for disabled and elderly people), privacy, and in information visualization. He is the editor of User Modeling and User-Adapted Interaction, editorial board member of World-Wide Web, Universal Access in the Information Society and Information Technology and Decision Making, and was the founding president of User Modeling Inc. Dr. Kobsa edited several books and authored numerous publications in the areas of user-adaptive systems, human-computer interaction and knowledge representation. He also co-founded a national workshop series and an international conference series in these areas.