The field of multi-agent systems and AI alignment is moving towards a deeper understanding of the complex interactions between human and AI agents. Researchers are exploring the economic tradeoffs between human and AI agents in bargaining games, highlighting the importance of evaluating not only the performance of AI agents but also the processes through which they negotiate. The development of mechanism design with outliers is also gaining traction, with studies revealing that discarding outliers can sometimes lead to counterintuitive outcomes. Furthermore, the evaluation of social capabilities in AI agents is becoming increasingly important, with a focus on assessing prosocial abilities and cooperation capabilities in complex environments. Noteworthy papers in this area include:

• Understanding Economic Tradeoffs Between Human and AI Agents in Bargaining Games, which compares humans, large language models, and Bayesian agents in a dynamic negotiation setting.
• Beyond the high score: Prosocial ability profiles of multi-agent populations, which applies a Bayesian approach to infer capability profiles of multi-agent systems and reveals underlying prosocial abilities of agents.
• Emergent Alignment via Competition, which studies a strategic setting where a human user interacts with multiple differently misaligned AI agents and shows that strategic competition can yield outcomes comparable to interacting with a perfectly aligned model.