A Theory of Optimal Agenda Design

This paper formalizes the problem of designing optimal agendas for voting over finite alternative spaces, when voters are assumed to be "naive," i.e., they do not vote strategically. The class of agendas considered here is quite broad, and includes, as special cases, such methods as pairwise voting, sequential and elimination procedures, partitioning schemes, and all binary procedures. Given individual preferences over the basic alternative space, and various assumptions about how individuals choose between subsets of alternatives, one can then formalize the problem of designing agendas as a dynamic programming problem and solve for optimal agendas, i.e., agendas having either the highest probability of leading to a given alternative or having the highest expected utility to the agenda setter. Illustrations are given showing how the methods can be applied in specific examples.