UCLA

This thesis consists of three essays studying the theory of bargaining and learning dynamics of matching platforms. The first essay studies the role of optimism in non-cooperative bargaining, while the second essay explores how introducing bargaining incentives affect trust building process in international relations context. The final essay considers learning incentives of matching platforms that utilize their matching technology to exploit or explore the quality of their constituents.

The first essay asks a theoretic question: does exaggerated optimism benefit an agent in bargaining? The paper analyzes a two agent non-cooperative bargaining model to study if, and when, one has incentive to over-report his level of optimism. It modifies the complete information Rubinstein bargaining model to let players hold different beliefs about which player makes an offer. Defining optimism over one's perceived recognition probability, I find that an agent always ``envies" a more optimistic agent, and has incentive to play optimism as strategic posture to benefit. The second part of the chapter introduces an asymmetry of information to the game, letting an agent be of a ``more optimistic" type with some known probability. I find that the less optimistic type 1) pretends to be the more optimistic type---``play optimism"---if his probability of being more optimistic is high enough, 2) reveals his type before the more optimistic type would have settled, and 3) benefits more by playing optimism the higher the probability of extreme optimism is.

The second essay studies social encounters that involve both trust building and bargaining. We show that while bargaining interferes with trust building in the sense that fully informative signaling becomes impossible, bargaining alongside trust-building actually improves welfare when initial trust is low. In contrast to the current literature, we show that actors improve welfare by building trust more slowly. Thus, windows of opportunity to build trust must be seized to prevent significant declines in expected welfare. We also characterize the evolution of stakes that lead to the best outcomes. Our analysis explains why trust building is so much more difficult than the current literature implies and illuminates the opportunities that produce the best outcomes between adversaries with something to lose.

The third essay studies how platforms can utilize its pooling ability both to generate flow output and to discover good agents at the same time. In a simple model of two types in continuous time, the paper identifies an exploration-exploitation trade-off: by only matching good agents to each other, the platform may maximize flow output while sacrificing discovery of new good agents; on the other hand, by keeping an integrated pool, the platform maximizes learning rate while sacrificing the number of good matches. We find that the optimal matching policy is bang-bang from full integration--until the discovery ratio of good agents hits a certain threshold--to full segmentation thereafter to maximize flow payoffs. We also characterize how the threshold ratio responds to parameters of the model.