About Me

Hi! My name is Eunice Yiu. I am a postdoctoral scholar at UC Berkeley. I am affiliated with the Department of Psychology and Berkeley AI Research Lab.

I study how humans and AI systems build, generalize, and update models of the world, and what happens when those systems interact at scale with real people. My work combines behavioral experiments with controlled model evaluation to probe causal reasoning, interactive learning, and generalization in both humans and large multimodal models. I am increasingly focused on the societal consequences of deployed AI: how interaction patterns shape learning outcomes, how values get encoded and evolve through continuous human-AI interaction, and how to evaluate reasoning beyond surface outputs.

Active projects include: fine-tuning VLMs via iterative feedback loops in simulated environments to study how models acquire causal tool use compared to human developmental patterns (Google-BAIR); causal learning and active exploration in humans and LLMs (with Mila); effects of AI on human cognitive and behavioral health (with Brown ARIA).

2026 News

Jun 2026	I am co-organizing the Topics in Intelligence: World Models and Social Reasoning at Berkeley Simons Institute for the Theory of Computing.
Jun 2026	I am co-organizing the Humans of Generative AI workshop at CVPR.
May 2026	My paper on empowerment gain and causal model construction is out in Philosophical Transactions of the Royal Society A.
Apr 2026	I co-organized the Towards a Unified Account of Motivation in Development workshop at Cognitive Development Society Conference.
Mar 2026	I visited Princeton Natural and Artificial Minds Institute as a Diverse Intelligences Fellow.
Jan 2026	I gave a Keynote at the Foundations of Agentic Systems Theory Workshop, AAAI, Singapore.