What does it mean for a language model to “know” something—and how should it communicate uncertainty to the people wh o use it? In this talk\, Jacob Andreas\, Associate Professor of Electrical Engineering and Computer Science at MIT\, will explore new approaches to bu ilding language models that not only model the world but also model themsel ves.
Andreas will show how optimizing for coherence and calibration—b eyond accuracy alone—can produce models that are both more factually consis tent and more reliable in expressing confidence. These advances raise press ing questions for governance: How should large models present information? What standards should guide their expression of reliability or doubt?
Moderated by Josh Joseph\, the Berkman Klein Center’s Chief AI Scientist\, this conversation will situate cutting-edge technical work on belief and u ncertainty in language models within the wider debates about interpretabili ty\, trust\, and the responsible use of AI. Navigating computer science and the questions it raises for broader society\, the conversation will conclu de by examining the policy implications raised by these complex issues.
Jacob Andreas is an associate professo r at MIT in EECS and CSAIL. He did his PhD work at Berkeley\, where he was a member of the Berkeley NLP Group and the Berkeley AI Research Lab. His re search aims to understand the computational foundations of language learnin g\, and to build general-purpose intelligent systems that can communicate e ffectively with humans and learn from human guidance.
Josh Jo seph is Berkman Klein’s Chief AI Scientist\, the first role of its kind at BKC. Josh leads teams to explore measuring and controlling the age ncy of AI systems\, benchmarking these systems beyond measures of“intellige nce\,” and to help build the infrastructure for in-house AI research. In ad dition to being BKC’s Chief AI Scientist\, Josh currently holds appointment s as a Visiting Scientist at MIT and a Lecturer on Law at Harvard Law Schoo l.
CATEGORIES:Speaker/Panel LOCATION:Berkman Klein Multipurpose Room (Room 515) GEO:0.000000;0.000000 ORGANIZER;CN="Jessica Weaver":MAILTO:jweaver@law.harvard.edu URL;VALUE=URI:https://hls.harvard.edu/events/belief-uncertainty-and-truth-i n-language-models/ ATTACH;FMTTYPE=image/png:https://hls.harvard.edu/wp-content/uploads/2025/09/1Speaker_16x9.png END:VEVENT BEGIN:VTIMEZONE TZID:America/New_York BEGIN:DAYLIGHT TZOFFSETFROM:-0500 TZOFFSETTO:-0400 DTSTART:20250309T070000 TZNAME:EDT END:DAYLIGHT END:VTIMEZONE END:VCALENDAR