KIM Juho

Professor, School of Computing, KAIST

KIM Juho
Times of the Remarks 2026. 11. 05. 11:30-12:00
Title When AIs Debate: An Experiment in Cognitive Sovereignty

KIM Juho is an Professor in the School of Computing at KAIST and a human-AI interaction researcher who leads the KAIST Interaction Lab (KIXLAB). His research aims at AI that strengthens human thinking and judgment rather than replacing it, and he has designed online systems in which thousands of people learn, deliberate, and decide together, deploying them in the real world to test what works.

At SDF 2019 he presented StarryThoughts, a live system joined by over 1,200 citizens, showing that people with different views can in fact understand one another.

He holds a BS from Seoul National University, an MS from Stanford, and a PhD from MIT, has received more than twenty paper awards in human-computer interaction, and delivered a keynote at NeurIPS 2022. He currently serves on the UN Independent International Scientific Panel on AI and is a co-founder of SkillBench, a startup building solutions that measure and improve how AI affects the capabilities of people and organizations at work.

When AIs Debate: An Experiment in Cognitive Sovereignty

In the real world it is hard to hold even one public debate on a given issue, and practically impossible to run it again with different participants or different rules. This talk presents an experimental public sphere in which AI agents with diverse profiles, built from real demographic data, deliberate among themselves, allowing the same issue to be tested under thousands of conditions and combinations.

In this agent-based public sphere, perspectives that rarely surface in human debate can emerge, and we can observe both when deliberation succeeds and when it collapses; we also show a system that lets citizens look inside these debates and change the conditions themselves.

We take an honest look at what these experiments reveal about the possibilities and the limits of simulation technology. The talk closes by asking how agents and simulation could contribute positively to human public deliberation, and what a public sphere that protects people's sovereignty over their own thinking should look like in the age of AI.