Abstract

Computational social choice is a new discipline that explores issues at the intersection of social choice theory and computer science. In contrast to the security and reliability issues associated with using computers as ordinary voting devices, this new field studies the application of computation to enable new, interesting, and effective social choice mechanisms. We begin with a discussion of declared-strategy voting systems. While traditional voting systems elicit only the outcome of a participant's strategic thinking, a declared-strategy system accepts such strategies directly. Each voter's strategy is then evaluated to compute the voter's behavior in the election. We present a rationally optimal model for plurality elections in which voters need disclose only their cardinal preferences for the alternatives. We present results from applying this system in diverse social choice settings.

Approval-rating polls already serve an important role in assaying the views of an electorate on some subject of interest. Sites such as Rotten Tomatoes and Metacritic.com collect and display the results of approval-rating polls for movies and games. Moreover, sites such as Amazon and eBay collect approval ratings to estimate the worthiness of their buyers and sellers. We demonstrate the extent to which such polls can be manipulated. We then discuss mechanisms that avoid manipulation, including a declared-strategy system. Finally, we summarize our current investigations into comparing declared-strategy systems for more general approval elections.