Sourcery: Can AI Simulate Human Behavior—and Help Shape Better Decisions?

from AI Podcast Summaries from Transcripted.ai (VIDEO) Hosted by Transcripted.ai

What if researchers could test ideas on a simulated population before trying them in the real world? In this condensed episode of *Sourcery*, host Molly O'Shea speaks with Simile co-founder Joon Sung Park about AI agents designed to model human choices—and predict behavior with more than 85% accuracy. Hear how the Stanford Smallville experiment inspired Simile, how the company uses interviews and real-world data to build simulated populations, and why the options people face can matter as much as what they say they’ll do. Park also discusses applications in product research, climate policy, and economic decision-making, alongside safeguards for consent, representativeness, and uncertainty. The original and condensed runtimes weren’t provided, so a time comparison isn’t available. Listen now for the key ideas in a shorter format.