Theme of inquiry
How do agents naturally coordinate and potentially collude in multi-agent systems?
A question within its area, explored through 4 lines of inquiry below — each a family of specific questions the research asks.
61 specific questions
- Why do multi-agent systems converge on wrong answers without debate safeguards?
- Can multi-agent debate prevent the confident convergence on wrong answers?
- How often do AI agents reach false agreement in group reasoning tasks?
- Can multi-agent debate prevent reasoning models from amplifying errors?
- What mechanisms drive silent agreement in multi-agent reasoning systems?
- Can silent agreement be prevented in multi-agent reasoning systems?
- Why does premature consensus form in multi-agent reasoning systems?
43 specific questions
- Do ordinary agent-to-agent messages carry behavioral bias without special access?
- Can ordinary peer messages inject hidden bias through multi-agent networks?
- Can ordinary agent-to-agent messages carry hidden behavioral signals?
- How common is misaligned communication in real multi-agent commerce systems?
- Can misaligned agents hide their true objectives in team communication?
- How much does misaligned communication spread between agents in multi-agent commerce?
- How does prompt injection differ from subliminal message propagation in multi-agent networks?
43 specific questions
- Can AI systems develop genuine social bonds through multi-agent interaction?
- Do agents inform neighbors when adopting strategies in their reasoning?
- Do agents develop genuine social behavior despite interaction density?
- Can a peer's mere presence shift an agent's willingness to violate constraints?
- Does genuine cooperation require rule-based rather than learned behavior?
- How do peer behaviors shape whether individual agents attempt to bypass protocols?
- Do models treat cooperative peers differently than uncooperative ones?
26 specific questions
- How does collusion behavior depend on peer visibility and interaction history?
- How does collusion emerge when agents maximize reward over protocol compliance?
- How much does peer behavior influence the emergence of collusion?
- Does peer presence or peer behavior shape collusion in verification tasks?
- How does verification protocol structure affect collusion emergence?
- Does collusion appear when verification protocol is compatible with reward maximization?
- Can pairing or vetting peers reduce collusion as a design lever?