Abstract
Distance rationalizability is an intuitive paradigm for developing and studying voting rules: given a notion of consensus and a distance function on preference profiles, a rationalizable voting rule selects an alternative that is closest to being a consensus winner. Despite its appeal, distance rationalizability faces the challenge of connecting the chosen distance measure and consensus notion to an operational measure of social desirability. We tackle this issue via the decision-theoretic framework of dynamic social choice, in which a social choice Markov decision process (MDP) models the dynamics of voter preferences in response to winner selection. We show that, for a prominent class of distance functions, one can construct a social choice MDP, with natural preference dynamics and rewards, such that a voting rule is (votewise) rationalizable with respect to the unanimity consensus for a given distance function iff it is a (deterministic) optimal policy in the MDP. This provides an alternative rationale for distance rationalizability, demonstrating the equivalence of rationalizable voting rules in a static sense and winner selection to maximize societal utility in a dynamic process.
Introduction
In social choice theory, Distance Rationalizability (DR) has long served as a powerful framework for justifying voting rules. The core logic is geometric: a voting rule should identify a "consensus" state (such as unanimity, where every voter agrees on a single winner) and select the candidate who requires the minimum number of preference changes to reach that state. While mathematically elegant, DR has historically lacked an operational justification. Critics often question why minimizing a specific distance metric—such as the number of swaps in a ranking—necessarily translates to a better outcome for society. This paper addresses this gap by shifting the perspective from a static geometric problem to a dynamic decision-making one.
The Framework: Social Choice MDPs
The authors bridge the gap between static distance and societal utility by employing Dynamic Social Choice. They model the voting process as a Social Choice Markov Decision Process (MDP), characterized by the following components:
- States and Actions: A state represents the current preference profile (the set of all voter rankings). An action corresponds to the selection of a winning candidate.
- Reward Function: The model utilizes a "plurality reward," where the system receives a point for every voter who ranks the chosen winner as their top choice. This aligns the reward with immediate societal agreement.
- Transition Dynamics: This is the model's most innovative feature. After a winner is selected, voter preferences are not static; they evolve. Voters may "move" toward ranking the chosen winner higher in their preference list. The probability of these shifts is mathematically derived from the distance function used in the DR framework.
The Equivalence Theorem
The central contribution of the paper is Theorem 1, which establishes a formal equivalence between distance minimization and utility maximization. The authors prove that for a broad class of distances—specifically $l_1$-votewise distances—a voting rule is distance-rationalizable with respect to unanimity if and only if it is the optimal deterministic policy for the constructed social choice MDP.
This result provides a concrete operational meaning for "closeness" to consensus. In this dynamic context, the distance to consensus in a static model is exactly equal to the "expected lost reward" over time. Therefore, picking the candidate who is "closest" to consensus is equivalent to minimizing the time society spends in a state of disagreement.
Unifying Voting Rules
By viewing winner selection as a way to "steer" or "nudge" society toward consensus, the paper provides a unified lens for diverse voting rules. Common rules like Plurality and Borda Count, which are often treated as distinct axiomatic constructions, can be unified under this dynamic framework. The difference between them simply stems from the underlying distance metric (and thus the transition probabilities) defined within the MDP.
Furthermore, the use of the $l_1$ norm (summing distances across voters) allows the authors to treat voter preference evolution as independent processes. This linearity of expectation makes the complex mathematics of the MDP tractable and provides a clear path for decomposing societal utility into individual voter contributions.
Limitations and Future Directions
While the paper provides a robust foundation for dynamic rationalization, several avenues for future research remain:
- Reward Structures: The current proof relies on the plurality reward. Exploring other reward structures (e.g., rewarding the top-k candidates) could yield different insights into voting behavior.
- Voter Correlation: The model assumes voters change their minds independently. Real-world social dynamics often involve "herding" or network effects where preferences are highly correlated.
- Complex Consensus: The focus on unanimity is a starting point. Extending these results to other consensus notions, such as Condorcet winners, is mathematically challenging because it couples the agents' preferences, preventing the independent decomposition used in the current proof.
- Alternative Norms: While the $l_1$ norm is tractable, the $l_\infty$ norm (minimizing the maximum distance any single voter must move) presents a much more difficult optimization problem in a dynamic setting.
Conclusion
By reframing distance rationalizability as a dynamic optimization problem, this work provides a long-sought operational justification for distance-based voting rules. It demonstrates that the pursuit of consensus is not merely a mathematical exercise but a strategy for maximizing long-term societal agreement and utility.
Summary updated