FraudSociety: A Generative Agent-Based Society for the Financial-Fraud Economy
Summary
- Researchers introduce FraudSociety, an LLM-agent-based simulator that jointly models informational and financial dynamics in the financial-fraud economy.
- The architecture integrates a dynamic social environment, persistent financial ledgers, and a Need-Intent-Action decision-making model for civilians.
- Fraudsters adapt communication and financial actions based on individual interaction histories and ledger states, creating a closed feedback loop.
- The simulator successfully reproduces macro-level fraud patterns and micro-level loss concentration matching real-world statistics.
- Counter-fraud experiments demonstrate that personalized interventions can reduce the repeat-payment rate significantly from 49.4% to 10.9%.
Why it matters
FraudSociety provides AI researchers and financial institutions with a high-fidelity, agent-based testbed to model complex fraud ecosystems and rigorously evaluate anti-fraud intervention strategies.