Behavioral grounding
We study how observed actions, incentives, memory, social influence, and context can be represented without reducing people to static personas.
RESEARCH
From human evidence to testable possible worlds.
WORLDSIM
Our research focuses on representing behavior, interaction, uncertainty, and environments well enough to improve the next real-world experiment.
Multi-agent interaction
Individual choices change when actors influence one another. We examine propagation, coordination, conflict, and emergent group behavior across repeated runs.
Scenario validity
A simulation is useful only when teams can see why it produced an outcome. Our work emphasizes evidence provenance, sensitivity analysis, calibration, and explicit uncertainty.
- Trace evidence to model assumptions
- Compare controlled counterfactuals
- Validate against observable outcomes
Environment models
Products, markets, cities, policies, and organizations impose different constraints. We research reusable environment representations that preserve the conditions relevant to a decision.
Responsible use
We design for human oversight, contestability, and clear limits. High-impact decisions require domain review and should never rely on a simulation as their sole source of truth.