Strategies Improve Social Welfare: An Empirical Study of Strategic Voting in Social Networks

Abstract

In utilitarian social choice settings, agents have cardinal utilities over candidates, while for many reasons they only report their ordinal rankings for candidates. One possible way is to use voting systems that have access to the ordinal rankings, to maximize the sum of utility of all agents, i.e. social welfare. However, most work does not consider strategic play when using ordinal preferences as a proxy for cardinal utilities. In this paper, we present a new decision-making model of strategic voting to map cardinal utilities to votes under incomplete information. We consider an iterative setting, where agents will update their votes based on information observed from their neighbors in a social network. We compare our model with three existing models, one is a simple heuristic model and the other two are boundedly rational models. Results from extensive simulations on both random networks and a real social network show that although the strategic play is not frequent, sometimes rare, it does improve the social welfare of selected winners. We investigate the Duverger’s Law, showing that only the simple heuristic model that requires low cognitive efforts like human voters can replicate the law.