Fairgame
An AI agent is an autonomous system that processes information, retains memory, and adapts its behavior over time to achieve specific goals. These agents power a range of intelligent systems, including decision-making tools, virtual assistants, and interactive chat agents.
FAIRGAME (Framework for AI Agents Bias Recognition using Game Theory) is a flexible framework built to assess the cooperativeness of AI agents. By assigning distinct identities and personalities, FAIRGAME uses game-theoretic dynamics to observe how agents interact, make decisions, and align with predefined behavioral expectations.
Fairgame for chat agents
A major challenge for organizations and companies is evaluating chat agents before deployment. While static tests provide some insights, they fall short of capturing the unpredictability of real human interactions. FAIRGAME enables the simulation of conversations with hundreds of AI-generated users, each with unique identities, personalities, goals, and requests. At the end of each interaction, both the chat agent and the AI-simulated user must make decisions, each triggering rewards or penalties for the chat agent. This process quantifies how well the chat agent's behavior aligns with the user-defined expectations.
About the project
FAIRGAME was developed by the AI Readiness and Assessment (AIRA) research group, with the help of the University of Trento, Teesside University, and the University of Cambridge. Relevant scientific literature on this project:
Fairgame: a framework for ai agents bias recognition using game theory, Buscemi, Alessio, et al., arXiv preprint arXiv:2504.14325 (2025).
Media and responsible AI governance: a game-theoretic and LLM analysis, Balabanova, Nataliya, et al. arXiv preprint arXiv:2503.09858 (2025).
Do LLMs trust AI regulation? Emerging behaviour of game-theoretic LLM agents, Buscemi, Alessio, et al., arXiv preprint arXiv:2504.08640 (2025).
For further information, feel free to contact us:
Define your customer agent's identity
