(CANCELED) Between Model and World: Trust, Transparency, and Theory of Mind in AI Systems

Abstract: This talk works through a disciplinary survey of what constitutes trust and trustworthiness in human-machine learning interaction. In positing a framing of “trust,” the question of transparency or explainability directly surfaces. With the recent arrival of reasoning models, there seems to be a transparent view on the chain-of-thought in the process toward and output. And yet, what does “transparency” yield in understanding and lucid engagement? In this sense, the last turn of the talk addresses the concept of theory of mind in AI systems: what world view, if any, do machine learning systems incorporate? And if AI Systems imagine worlds, what do those worlds look like.

Bio: Dr. Beth Coleman is director of the Knowledge Media Design Institute (KMDI), a multidisciplinary research institute supported by the Faculty of Information at the University of Toronto. An Associate Professor of Data & Cities at the Institute of Communication, Culture, Information and Technology and Faculty of Information, Coleman works in the disciplines of science and technology studies and generative aesthetics with research focuses on artificial intelligence & smart technology, urban data and civic engagement, and transmedia arts. She is the author of Hello Avatar (MIT press) and Reality Was Whatever Happened: Octavia Butler AI and Other Possible Worlds (K. Verlag). She has been a Google Brain and Responsible AI senior visiting researcher as well as a 2021 Google Artists and Machines Intelligence awardee. She is a research lead on AI policy and praxis at the Schwartz Reisman Institute for Technology & Society. Her research affiliations have included the Berkman Klein Center for Internet & Society, Harvard University; Microsoft Research New England; Data & Society Institute, New York; and expert consultant for the European Commission Digital Futures. She is currently working on the monograph, AI in the World: Perils and Possibilities of a General Purpose Technology.