CLR (cooperation / conflict)
How do multi-agent failure modes behave under strategic pressure—especially when cooperation breaks down?
Introduction
CLR researches cooperation under AI competition, s-risks, and multipolar failure modes, including Cooperative AI Foundation work. Its conflict-and-cooperation framing asks how multi-agent dynamics behave when cooperation breaks down under strategic pressure. The agenda engages Goodhart Selection and Acausal Coordination as measurement cousins to narrative multipolar stories.
Who carries it: Center on Long-Term Risk
What they aim to do. Reduce worst-case outcomes that arise when AI systems compete or come into conflict with one another.
The hard question. How do multi-agent failure modes behave under strategic pressure—especially when cooperation breaks down?
What they produce. Research and grantmaking on cooperation under AI competition, s-risks, and multipolar failure modes, including Cooperative AI Foundation work.
Key terms. Key terms include cooperation, conflict, s-risks, multipolar failure, and CAIF (Cooperative AI Foundation).
Related field cruxes. Goodhart Selection; Acausal Coordination
What they contribute. Conflict-and-cooperation framing for alignment research, plus Cooperative AI Foundation work on building cooperative AI under competitive pressure.
How this project treats it. Typed measurement of Goodhart Selection and Acausal Coordination differs from narrative multipolar stories that lack the same measurement discipline.
Links
- CLR
- Critch & Krueger 2020 — ARCHES
- Dafoe et al. 2020 — Open Problems in Cooperative AI
- Cooperative AI Foundation (CAIF)
See the coverage matrix for evidence tagged to this agenda, and the glossary for shared terms.