The Foundations of Cooperative AI Lab (FOCAL) is a Carnegie Mellon research group directed by Vincent Conitzer, Professor of Computer Science at CMU with affiliate appointments in Machine Learning, Philosophy, and the Tepper School of Business. Conitzer additionally serves as Head of Technical AI Engagement at the University of Oxford's Institute for Ethics in AI. FOCAL's named research directions are designing preferences, beliefs, and identities for AI; open-source game theory and multilateral commitments; foundations of multi-agent learning; decision-theoretic foundations for game theory; and self-locating beliefs (the Sleeping Beauty / anthropic-reasoning end of decision theory).
The Cooperative AI Foundation grant of $500,000 over the 2021 to 2025 period supported the lab's founding. CAIF's grant summaries list FOCAL among its founding-period investments. A separate $3 million gift to CMU (per CAIF's grants page) provides additional institutional support, making the CAIF grant a co-funding contribution rather than the sole source.
Recent FOCAL output includes papers co-authored by Conitzer with Caspar Oesterheld, Maxime Riche, Filip Sondej, and Jesse Clifton on surrogate goals for safer bargaining between LLM-based agents, and a body of work on commitment, threat, and bargaining dynamics in multi-agent settings. The lab also runs a workshop series (FOCAL at AAAI) and an online seminar series hosted by CAIF that brings together researchers working on multi-agent cooperation, mechanism design, and decision theory relevant to autonomous AI agents.
Within the open-source AI stack the work is at the agents layer and the evaluation meta-layer. As multi-agent systems and agent-to-agent protocols (MCP, A2A) become production patterns, FOCAL's outputs are part of the small set of academic groups producing theoretical foundations for what happens when independent AI agents negotiate, commit, threaten, or coordinate. The work is upstream of practical safety guardrails for agent-to-agent interactions.
Recipient
Vincent Conitzer (Carnegie Mellon FOCAL lab)
Funder
Cooperative AI Foundation · foundation · UK
Funds research that improves AI agents' capacity for cooperation with each other and with humans, including measurement of cooperation-relevant capabilities and propensities.