Agentmandering: A Game-Theoretic Framework for Fair Redistricting via Large Language Model Agents

2025-11-06

AI Summary

Existing redistricting methods prioritize generating legally compliant maps but overlook strategic manipulation by partisan actors during map selection, leading to implicit bias. This paper introduces Agentmandering, the first framework to model redistricting as a turn-based game between two adversarial political-interest LLM agents, implementing a progressive partitioning process via a game-theoretic Choose-and-Freeze protocol—where agents alternately select and freeze districts. The approach ensures interpretability, stability, and fairness. Evaluated on 2020 U.S. Census data across all states, Agentmandering significantly reduces partisan bias: unfairness variance drops by two to three orders of magnitude relative to baseline methods, with particularly strong robustness in swing states. Its core contribution lies in explicitly incorporating strategic interaction into redistricting design, thereby mitigating partisan skew at its source—i.e., the map-selection process itself.

Technology Category

• Game Theory and Economic Paradigms: Fair Division
• Multiagent Systems: Adversarial Agents
• Knowledge Representation and Reasoning: Preferences

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Redistricting plays a central role in shaping how votes are translated into political power. While existing computational methods primarily aim to generate large ensembles of legally valid districting plans, they often neglect the strategic dynamics involved in the selection process. This oversight creates opportunities for partisan actors to cherry-pick maps that, while technically compliant, are politically advantageous. Simply satisfying formal constraints does not ensure fairness when the selection process itself can be manipulated. We propose extbf{Agentmandering}, a framework that reimagines redistricting as a turn-based negotiation between two agents representing opposing political interests. Drawing inspiration from game-theoretic ideas, particularly the extit{Choose-and-Freeze} protocol, our method embeds strategic interaction into the redistricting process via large language model (LLM) agents. Agents alternate between selecting and freezing districts from a small set of candidate maps, gradually partitioning the state through constrained and interpretable choices. Evaluation on post-2020 U.S. Census data across all states shows that Agentmandering significantly reduces partisan bias and unfairness, while achieving 2 to 3 orders of magnitude lower variance than standard baselines. These results demonstrate both fairness and stability, especially in swing-state scenarios. Our code is available at https://github.com/Lihaogx/AgentMandering.

Problem

• Addresses partisan manipulation in redistricting through strategic agent negotiation
• Reduces bias by modeling redistricting as turn-based game-theoretic process
• Ensures fairness in map selection using LLM agents with constrained choices

Innovation

• Uses turn-based negotiation between opposing agents
• Applies game-theoretic Choose-and-Freeze protocol
• Leverages LLM agents for strategic redistricting decisions