Today, we are getting acquainted with autonomous computers and AI systems that make decisions for us. Yet, they are becoming more system-like, engaging in conversations and making decisions with other machines, and those decisions affect our lives. My work explores the design and function of these systems.
I study how autonomous agents make decentralized decisions and how those decisions aggregate into system-level performance, in networks that are open (agents enter and leave), coevolving (both agent behavior and network structure change together over time), and noisy (errors and imperfect signals are the norm, not the exception).
Agents pursue a mix of shared and private objectives, hold heterogeneous and incomplete knowledge, and operate under bounded rationality — there is no central controller and, depending on the setting, no guaranteed shared knowledge pool. Because information itself is costly to acquire, maintain, and verify, I treat knowledge as a scarce resource subject to the same budget constraints as time or money. This scarcity has consequences beyond any single agent’s decisions: when agents negotiate under a mix of shared and private objectives with no central coordinator, the resulting system-level outcome can fall measurably short of what coordinated behavior would achieve, a gap I study as a price of anarchy, and the negotiation process itself, not just the information it exchanges, carries its own transaction costs of search, bargaining, and enforcement. My interest is in the rules and boundaries that emerge or are imposed on this kind of system, and how those rules, together with these economic frictions, shape both individual behavior and aggregate outcomes.al behavior and aggregate outcomes.
Full Research PDFs are available on request.