Recent technological developments have enabled collecting and processing valuable information, which has greatly influenced decision-making in various areas. These advances have motivated the study of effective information provision strategies. In this thesis, we study the role of information provision in two contexts: online rating systems and traffic systems. In both systems we identify paradoxical situations in which providing more information to users leads to a worse performance of the system.
In the first part, we develop a model of Bayesian learning from online reviews and investigate the conditions for asymptotic learning of the quality of a product. We then characterize the speed of learning under different rating systems and show that the incentives of the platform are aligned with maximizing the speed of learning. Finally, we study how the platform can design the rating systems (in terms of information collection and information provision schemes) to accelerate learning. In particular, we identify situations in which providing more information leads to slower learning.
In the second part, we develop a framework for systematically analyzing how changes in the information sets of users in a traffic network (e.g., due to route guidance systems) impact the traffic equilibrium, and show the conditions under which even those with access to additional information may suffer greater congestion. We formulate this problem in the form of Informational Braess' Paradox (IBP), which extends the classic Braess' Paradox in congestion games. IBP asks whether users receiving additional information can become worse off. We provide a comprehensive answer to this question by providing a tight characterization of network topologies under which IBP emerges.
Prof. Asu Ozdaglar
Prof. Daron Acemoglu
Prof. John Tsitsiklis