Portrait of Anja Thieme

Anja Thieme

Senior Researcher

About

I’m a human-computer interaction (HCI) researcher in the Biomedical Imaging team within Microsoft Health Futures, based at Microsoft Research Cambridge, UK. My research explores innovative and responsible approaches for applying AI and Machine Learning in real-world healthcare contexts. Working in inter-disciplinary teams, I follow a human-centered approach to the end-to-end development of novel AI applications and systems. Specifically, my work investigates AI use in radiology workflows to improve patient outcomes through the development of clinically useful, interpretable and actionable ML insights.

More broadly, this work enables me to define strategy and processes that ensure AI technology is developed safely and responsibly. Working as part of Microsoft’s Aether working groups and with members of the Office of Responsible AI (ORA), a key focus of my work is the development of tools and other resources to drive forward the adoption of responsible AI practices; to advance public understanding, and to inform AI policy. This includes educational materials to appropriately communicate AI capabilities and limitations to end-users, and for considering wider socio-technical implications of AI use.

For more up-to-date information, please visit my homepage: https://designandwellbeing.com/

Past projects at Microsoft include:

  • Project Talia, which investigated the use of ML models to improve the quality of care delivered through online psychotherapy programs.
  • Project Tokyo that developed an interactive, computer-vision based system to extend the capabilities of people who live with vision impairments.
  • Project Torino, and its commercialized counterpart Code Jumper, a collaborative learning environment for teaching children how to code independent of their visual abilities.
  • Disruptive Displays, a research project that explored alternative approaches to designing and configuring digital displays (as a material that can be cut, folded, and externally imaged).