Influence Benchmarks for AI Systems
Summary
The Department of War, via DARPA, seeks proposals to develop a simulated market testbed that can benchmark and characterize bias, deception, and collaboration behaviors of AI agents, especially large‑language models, in dynamic economic environments. The effort aims to create a sandbox for evaluating AI decision making against human behavioral baselines to ensure safe integration of AI into high‑stakes warfighting contexts.