Stanford computer science professor Michael Bernstein discusses how AI agents can be built to model and simulate human behavior, drawing on his lab's research in human-AI interaction and social computing. Delivered as a Stanford Online webinar in collaboration with Global Alumni, the talk covers how large language model agents can represent individual behaviors and social dynamics, what such simulations are useful for, and where they break down or mislead. Bernstein frames the discussion around the design challenges this raises for builders of intelligent systems, including how to evaluate whether an agent's simulated behavior is a faithful stand-in for a real person's, and what responsibilities come with deploying systems that claim to model people. The session runs an hour and stays close to Bernstein's own published research on generative agents and social computing rather than general AI commentary, making it a reasonable primer on where this subfield stands.