Prediction markets are markets designed for information aggregation. To achieve its goal, a prediction market offers a contract whose future payoff is tied to outcomes of an event of interest. Participants of the market express their information about the event through trading the contract. The market price hence potentially incorporates the information of all participants and approximately represents a consensus forecast for the event. Prediction markets have shown great potential as highly effective information aggregation tools in practice. However, they may be subject to manipulation. Participants have incentives to lie about their information in order to seize more profit in the market or be rewarded outside the market. With manipulation, information aggregation may fail and the credibility of market prices is put into question. Another challenge is the computational problem of operating combinatorial prediction markets. When the mechanism becomes more expressive and participants can reveal their information on combinations of outcomes, the process of operating a market may become computationally intractable, which makes such combinatorial prediction markets impractical to use. The goal of this project is to perform a foundational study on issues that arise in prediction market design because of the self-interest of participants and the need of expressiveness. Using computer science theory and game theory as the main research approaches, Dr. Chen will establish the strategic and computational foundations for prediction markets so as to design market mechanisms that are not only theoretically sound but also practically applicable for information aggregation in complex real-world settings.
This project is centered around two themes: manipulation-resistant prediction markets and computationally efficient combinatorial prediction markets. The research on manipulation-resistant prediction markets investigates when manipulation arises, how manipulation affects information aggregation, and how to design prediction market mechanisms to explicitly control the impact of manipulation. It will provide answers to the most fundamental question of whether people can trust the information conveyed by prediction markets and design robust market mechanisms for information aggregation. The research on computationally efficient combinatorial prediction markets examines the tradeoff between expressiveness and computational complexity of combinatorial prediction markets, designs tractable combinatorial prediction markets, and develops efficient approximation algorithms for intractable combinatorial prediction markets. It will help to transfer combinatorial prediction markets from theoretical artifacts to practical devices for aggregating richer information. This project does not depend on any specific application domain. Hence, mechanisms designed in this project can be applied to aggregate information in many settings, including business, government, and society. Results of this project will also help to inform policy making on the possible regulation of prediction markets.
