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

TL;DR

Agentmandering reframes redistricting as a strategic negotiation between two LLM-powered agents, implementing the Choose-and-Freeze protocol to curb partisan manipulation while drawing districts. By coupling a constrained candidate-generation step with adversarial freezing and leveraging state-specific political profiles, the framework yields procedurally transparent maps that are robust to strategic selection. Empirical results on post-2020 U.S. Census data show 2–3 orders of magnitude lower variance and improvements in partisan bias and unfairness, with particularly strong performance in swing states, and demonstrate robustness across diverse LLMs and generator choices. The approach offers a scalable, interpretable alternative to traditional ensemble methods, bridging game-theoretic fairness with practical redistricting tooling, and the code is publicly available.

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 \textbf{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 \textit{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.

Figures (7)

• Figure 1: (a) Gerrymandering example showing district manipulation. (b) Correlation analysis of four redistricting metrics and comparison of two districting plans.
• Figure 2: The Agentmandering framework.
• Figure 3: Normalized distribution of Unfairness across seven states for Agentmandering and Recom.
• Figure 4: Comparison of LLM-based and rule-based agent decisions in terms of Unfairness.
• Figure 5: Distribution of Unfairness scores across additional U.S. states, comparing Agentmandering and Recom. Lower scores indicate better voter satisfaction.