← All research

Research overview

Can a synthetic society forecast collective behaviour?

2026 · Research system · historical backtesting in progress

An experiment in using populations of AI agents to model collective behaviour rather than asking a single model for a prediction. Synthetic individuals are given different attributes, information environments and decision contexts, then allowed to form independent choices whose aggregate signal can be tested against real outcomes. The broader question is whether agent-based societies can generate predictive information that simpler forecasting methods miss.

Research focus

  • Multi-agent systems
  • Synthetic populations
  • Collective behaviour
  • Forecasting
  • Out-of-sample validation
  • Agent-based modelling
  • Artificial intelligence
  • Decision systems

Status

Research system · historical backtesting in progress. This page provides a concise overview of the work and its research domain.