Models that assume people don't change
The mathematical models used to forecast epidemics, climate-policy shifts, and other societal challenges mostly treat human behavior as static — yet behavior is what drives real-world outcomes. COVID-19 made the gap vivid: shifting risk perception, fatigue, and polarization repeatedly broke models that assumed people don't change.
A behavioral layer for computational models
Sociodyne is a generalizable behavioral layer for computational models. We couple established mathematical models — SIR-type and compartmental epidemic models, and system-dynamics simulations — with AI agents that simulate human behavior from real demographic and cognitive characteristics, run through an AI-driven modeling workflow.
The result: models that capture how people actually respond, producing sharper forecasts and clearer guidance on how to time and target interventions.
Behavior at scale is finally practical
Behavior is the acknowledged missing piece in these models, and modern AI finally makes it practical to represent rich, heterogeneous behavior at scale — and to operate complex modeling pipelines end to end.
Where the engine goes
We begin with infectious-disease forecasting and intervention design, then extend the same engine to climate, public health, and other behavior-driven problems.
- Public-health agencies
- Pharmaceutical & vaccine developers
- Health systems
- Insurers & reinsurers
Ways to work with us
Platform subscriptions, licensing of the behavioral engine and API, and custom modeling and decision-support engagements.
Research foundation
Built on an active NSF-funded research program and peer-reviewed publications in cognitive and behavioral epidemiological modeling, developed at the University of Vermont.
A public-benefit company
Sociodyne is a Vermont benefit corporation. We are developing our platform in partnership with the University of Vermont, supported by federal research funding, and validating it on a first infectious-disease use case.