On the Development of Voter Transition Models for Social Choice Markov Decision Processes

Social Choice Markov Decision Processes (SCMDP) have recently been proposed for use in online advocacy groups as a means for optimizing the voting process. In our paper, we extend the scope of this approach to the political context. We argue that SCMDPs can be useful tools for political campaigns and other initiatives to increase candidate popularity and satisfaction among constituents. We propose a survey-based methodology to construct a parsimonious voter transition model, which in turn provides the experimental foundation for the optimizing SCMDP. Based on hypothetical survey data, we have parameterized various voter transition models and implemented corresponding SCMDPs. The overall approach was validated through sensitivity analysis and simulation. Our results clearly show that decision processes in the political arena can benefit from this approach, especially when voters behave dynamically with respect to a policy maker\´s actions. A dynamic decision policy is noticeably better than a more "simplistic" scheme such as pure majority-based election.