pith. sign in

arxiv: 2604.19985 · v1 · submitted 2026-04-21 · 💻 cs.GT

Geometric Comparisons of Electoral Rules Under Feedback

Pith reviewed 2026-05-10 00:35 UTC · model grok-4.3

classification 💻 cs.GT
keywords electoral rulespolarization dynamicswinner radiussupporter centroid radiusfeedback adaptationgeometric primitivesvoting simulationcontraction bounds
0
0 comments X

The pith

Electoral rules create a tradeoff between shrinking voter disagreement and candidate dispersion through two geometric radii under repeated feedback.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces the winner radius R_t as the maximum distance from the election winner to any voter and the supporter centroid radius S_t as the maximum distance from a candidate to its support centroid. It shows that R_t bounds how much voter positions can contract in one step while S_t bounds candidate dispersion, yet the two radii pull in opposite directions. Rules that keep R_t small tend to enlarge S_t and vice versa, and placing the winner at the voter median does not remove the tension because the median and the Chebyshev center are distinct. Simulations of seven standard rules plus a convex benchmark across many electorate profiles confirm that winner-take-all systems produce small S_t but large R_t and slower voter depolarization, while mixtures reverse the pattern. The work further separates the per-step goal of minimizing R_t from the goal of minimizing disagreement directly, showing they yield different long-run trajectories for both voters and candidates.

Core claim

We show that R_t controls a one-step contraction bound on voter disagreement and S_t plays the analogous role for candidate dispersion, and that these two objectives are in tension. Rules that reduce R_t tend to increase S_t, and vice versa. A winner close to the voter median does not resolve the tension, since proximity to the median and proximity to the Chebyshev center are different objectives. The empirical results confirm the theoretical tradeoff: winner-take-all rules achieve small S_t at the cost of large R_t and weaker voter depolarization, while convex-combination rules reverse this. Minimizing R_t per step and minimizing voter disagreement per step are distinct objectives with不同长程的

What carries the argument

The winner radius R_t (maximum distance from winner to farthest voter) and supporter centroid radius S_t (maximum distance from candidate to its support centroid), which separately bound contraction rates for voter disagreement and candidate dispersion under adaptive updates.

If this is right

  • Winner-take-all rules achieve small S_t but large R_t and weaker voter depolarization.
  • Convex-combination rules achieve the opposite pattern of small R_t but large S_t.
  • Proximity of the winner to the voter median does not guarantee small R_t because the median differs from the Chebyshev center.
  • Minimizing R_t at each step produces different long-run voter and candidate trajectories than directly minimizing disagreement at each step.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Voting-system designers could target explicit control of both radii when constructing hybrid rules rather than optimizing one in isolation.
  • If real electorates adapt roughly as modeled, proportional or mixed rules may reduce sustained voter polarization more effectively than pure winner-take-all systems.
  • The median-versus-Chebyshev distinction suggests that worst-case voter coverage, not average proximity, is the relevant metric for depolarization speed.
  • The framework could be extended to test whether periodic switching between rule types breaks the accumulated tension between the two radii.

Load-bearing premise

The modeled update rules for how voters and candidates shift positions after each election outcome capture the actual dynamics without strategic manipulation or unmodeled external factors.

What would settle it

A simulation or real-world election sequence in which a rule that consistently produces smaller R_t also produces smaller S_t, or in which the predicted one-step contraction bounds on disagreement fail to match observed changes.

Figures

Figures reproduced from arXiv: 2604.19985 by Sumit Mukherjee.

Figure 1
Figure 1. Figure 1: Median difference from Plurality with interquartile bands across 1134 runs: voter [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Average start-to-end change in voter pairwise distance by system and mechanism [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Mean voter pairwise distance change by system and camp-balance setting. Plurality’s [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: reports end-state St under the µ = 0 supplementary grid. For rules with smooth assignment, this is the regime where Proposition 3.10 applies; Plurality is reported descriptively, since Assumption 3.9 fails under hard Voronoi assignment (see the Scope remark in Section 3). The ranking is the reverse of the Rt ranking: rules with hard assignment produce the smallest end-state St (Plurality: 0.083 under Stati… view at source ↗
Figure 5
Figure 5. Figure 5: Mean start-to-end change in voter pairwise distance versus mean change in winner [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Oracle comparison across 24 replicates. Shaded bands are interquartile ranges. [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
read the original abstract

We study how electoral rules shape polarization dynamics when voters and candidates both adapt to repeated election outcomes. We introduce two geometric primitives for comparing rules under this feedback: the \emph{winner radius} $R_t = \max_i \|x_i - w^{(t)}\|$, the distance from the winner to the farthest voter, and the \emph{supporter centroid radius} $S_t = \max_j \|c_j - s_j^{(t)}\|$, the largest gap between any candidate and their support base. We show that $R_t$ controls a one-step contraction bound on voter disagreement and $S_t$ plays the analogous role for candidate dispersion, and that these two objectives are in tension. Rules that reduce $R_t$ tend to increase $S_t$, and vice versa. A winner close to the voter median does not resolve the tension, since proximity to the median and proximity to the Chebyshev center are different objectives. We use this framing to organize a simulation study across seven standard electoral rules and one convex-combination benchmark, comprising 1000+ runs across diverse electorate profiles, voter mechanisms, and camp-balance settings. The empirical results confirm the theoretical tradeoff: winner-take-all rules achieve small $S_t$ at the cost of large $R_t$ and weaker voter depolarization, while convex-combination rules reverse this. An oracle comparison further shows that minimizing $R_t$ per step and minimizing voter disagreement per step are distinct objectives with different long-run consequences for both voter and candidate dynamics.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 3 minor

Summary. The manuscript introduces two geometric primitives—the winner radius R_t = max_i ||x_i - w^(t)|| and the supporter centroid radius S_t = max_j ||c_j - s_j^(t)||—to compare electoral rules in a repeated-election setting where both voters and candidates adapt their positions based on outcomes. It claims that R_t supplies a one-step contraction bound on voter disagreement while S_t does likewise for candidate dispersion, that these objectives are in tension (rules reducing one tend to increase the other), and that proximity to the voter median does not resolve the tension because it differs from proximity to the Chebyshev center. These relations are used to organize an extensive simulation study (1000+ runs) across seven standard rules plus a convex-combination benchmark, under varied electorate profiles, adaptation mechanisms, and camp-balance conditions; the simulations are said to confirm the theoretical tradeoff, with winner-take-all rules achieving small S_t at the cost of large R_t and weaker long-run voter depolarization, while convex combinations reverse the pattern. An oracle comparison further shows that per-step minimization of R_t and direct minimization of voter disagreement are distinct objectives with different long-run consequences.

Significance. If the one-step bounds and the simulation evidence hold under the stated adaptation rules, the work supplies a clean geometric vocabulary for analyzing polarization dynamics under feedback, separating the effects of winner selection from support-base alignment. The combination of explicit contraction statements with a large-scale numerical campaign (multiple rules, profiles, and mechanisms) and an oracle baseline is a strength; it moves beyond static spatial models toward dynamic comparisons that could inform institutional design. The distinction between median proximity and Chebyshev-center proximity is a useful clarification that stands independently of the simulations.

major comments (2)
  1. [§3 (contraction statements) and §5 (long-run claims)] The central theoretical claim is that R_t controls a one-step contraction on voter disagreement and S_t the analogous contraction on candidate dispersion, with the two objectives in tension. However, because the winner at step t+1 is chosen from the updated candidate positions (which evolve according to S_t) and voter updates depend on that winner, the iterated contraction factor is not controlled solely by the initial R_t. No joint fixed-point analysis, multi-step product bound, or inductive argument is supplied showing that the product of the per-step factors remains bounded away from 1 under simultaneous adaptation; consequently the long-run depolarization ordering asserted in the abstract and §5 rests on the simulation evidence rather than on the geometric primitives alone.
  2. [§2 (model) and §3 (bounds)] The adaptation equations for voters (how they update toward the winner) and candidates (how they update toward their supporters) are invoked to derive the contraction bounds, yet the manuscript does not display the exact update maps or the Lipschitz constants used in the proofs. Without these, it is impossible to verify that the stated one-step factors are tight or that they survive the coupling; an appendix containing the precise maps and the derivation of the contraction constants would be required to make the theoretical claims self-contained.
minor comments (3)
  1. [§2] Notation for the voter and candidate position vectors (x_i, c_j) and the time indices on the radii should be introduced once in §2 and used consistently; occasional reuse of w without the superscript (t) creates ambiguity when comparing across rules.
  2. [§5] The simulation section reports aggregate trends but does not tabulate the per-rule mean and standard deviation of final voter disagreement and candidate dispersion; adding such a table would make the empirical confirmation of the R_t–S_t tradeoff easier to assess at a glance.
  3. [§5] The oracle comparison is described as showing that minimizing R_t per step differs from minimizing disagreement per step, but the precise definition of the oracle (whether it has access to future positions or only current geometry) is not stated; a one-sentence clarification would remove ambiguity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful and constructive review. The comments highlight important aspects of the theoretical claims that require clarification and additional detail. We respond to each major comment below and indicate the revisions we plan to implement.

read point-by-point responses
  1. Referee: The central theoretical claim is that R_t controls a one-step contraction on voter disagreement and S_t the analogous contraction on candidate dispersion, with the two objectives in tension. However, because the winner at step t+1 is chosen from the updated candidate positions (which evolve according to S_t) and voter updates depend on that winner, the iterated contraction factor is not controlled solely by the initial R_t. No joint fixed-point analysis, multi-step product bound, or inductive argument is supplied showing that the product of the per-step factors remains bounded away from 1 under simultaneous adaptation; consequently the long-run depolarization ordering asserted in the abstract and §5 rests on the simulation evidence rather than on the geometric primitives alone.

    Authors: We acknowledge the validity of this observation. The one-step contraction bounds are derived for fixed positions at each time step and provide a geometric explanation for the per-step tradeoff between voter depolarization and candidate dispersion. However, due to the coupling between winner selection and the evolving positions, a rigorous multi-step bound or fixed-point analysis is not developed in the current manuscript. Consequently, the long-run depolarization ordering is supported by the simulation results rather than purely by the theoretical primitives. In the revised version, we will add a clarifying discussion in Section 5 to explicitly note this limitation and emphasize that the simulations (over 1000 runs across varied conditions) provide empirical confirmation of the tradeoff and ordering. We will also mention that deriving a joint inductive bound remains an interesting direction for future research. revision: partial

  2. Referee: The adaptation equations for voters (how they update toward the winner) and candidates (how they update toward their supporters) are invoked to derive the contraction bounds, yet the manuscript does not display the exact update maps or the Lipschitz constants used in the proofs. Without these, it is impossible to verify that the stated one-step factors are tight or that they survive the coupling; an appendix containing the precise maps and the derivation of the contraction constants would be required to make the theoretical claims self-contained.

    Authors: We agree that the manuscript would benefit from greater explicitness regarding the adaptation mechanisms. In the revision, we will append a new section or appendix that details the exact update equations for both voters and candidates, specifies the relevant Lipschitz constants, and walks through the derivations of the contraction bounds for R_t and S_t. This will make the theoretical claims fully self-contained and verifiable. revision: yes

Circularity Check

0 steps flagged

No significant circularity; geometric primitives yield independent contraction bounds

full rationale

The paper defines new geometric quantities R_t and S_t, then derives one-step contraction bounds on voter disagreement and candidate dispersion directly from the geometry of the winner and supporter centroids. These bounds follow from the definitions via standard contraction-mapping arguments rather than by re-labeling fitted parameters or importing unverified self-citations. The claimed tension between minimizing R_t and S_t is likewise a direct geometric observation, not a tautology. Long-run behavior is supported by separate simulation evidence rather than asserted as a theorem, so the derivation chain remains self-contained and does not reduce to its inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 2 invented entities

Central claim rests on definitions of R_t and S_t as controls on contraction, plus domain assumptions about voter/candidate adaptation mechanisms and electorate profiles; no explicit free parameters or invented entities beyond the two radii as measurement tools.

axioms (2)
  • domain assumption Voters and candidates update positions based on repeated election outcomes according to specified mechanisms
    Core modeling choice enabling the feedback dynamics and bounds.
  • domain assumption The geometric distances R_t and S_t capture the primary drivers of disagreement and dispersion
    Assumption that these primitives suffice without missing strategic or higher-order effects.
invented entities (2)
  • winner radius R_t no independent evidence
    purpose: Measure to bound one-step voter disagreement contraction
    New geometric primitive defined as max distance from winner to any voter.
  • supporter centroid radius S_t no independent evidence
    purpose: Measure to bound candidate dispersion
    New geometric primitive defined as max gap between candidate and their support centroid.

pith-pipeline@v0.9.0 · 5561 in / 1451 out tokens · 29478 ms · 2026-05-10T00:35:36.442793+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

18 extracted references · 18 canonical work pages

  1. [1]

    Proceedings of the Twenty-Ninth

    Anshelevich, Elliot and Bhardwaj, Onkar and Postl, John , title =. Proceedings of the Twenty-Ninth. 2015 , publisher =

  2. [2]

    Gkatzelis, Vasilis and Halpern, Daniel and Shah, Nisarg , title =. 2020. 2020 , publisher =

  3. [3]

    and Munagala, Kamesh , title =

    Goel, Ashish and Krishnaswamy, Anilesh K. and Munagala, Kamesh , title =. Proceedings of the 2017. 2017 , publisher =

  4. [4]

    arXiv preprint arXiv:2111.03694 , year =

    Charikar, Moses and Ramakrishnan, Prasanna , title =. arXiv preprint arXiv:2111.03694 , year =

  5. [5]

    British Journal of Political Science , volume =

    Diermeier, Daniel and Li, Christopher , title =. British Journal of Political Science , volume =. 2023 , doi =

  6. [6]

    , title =

    Meyn, Sean and Tweedie, Richard L. , title =. 2009 , publisher =

  7. [7]

    1957 , publisher =

    Downs, Anthony , title =. 1957 , publisher =

  8. [8]

    and Hinich, Melvin J

    Enelow, James M. and Hinich, Melvin J. , title =. 1984 , publisher =

  9. [9]

    , title =

    Cox, Gary W. , title =. American Journal of Political Science , volume =

  10. [10]

    Journal of Political Economy , volume =

    Callander, Steven and Carbajal, Juan Carlos , title =. Journal of Political Economy , volume =. 2022 , doi =

  11. [11]

    , title =

    DeGroot, Morris H. , title =. Journal of the American Statistical Association , volume =. 1974 , doi =

  12. [12]

    Journal of Artificial Societies and Social Simulation , volume =

    Hegselmann, Rainer and Krause, Ulrich , title =. Journal of Artificial Societies and Social Simulation , volume =

  13. [13]

    Models of Social Influence: Towards the Next Frontiers , journal =

    Flache, Andreas and M. Models of Social Influence: Towards the Next Frontiers , journal =. 2017 , doi =

  14. [14]

    and Lipsitz, Keena and Bizyaeva, Anastasia and Franci, Alessio and Lelkes, Yphtach , title =

    Leonard, Naomi E. and Lipsitz, Keena and Bizyaeva, Anastasia and Franci, Alessio and Lelkes, Yphtach , title =. Proceedings of the National Academy of Sciences , volume =. 2021 , doi =

  15. [15]

    and Duggan, John , title =

    Banks, Jeffrey S. and Duggan, John , title =. Caltech Social Science Working Paper , number =

  16. [16]

    International Journal of Game Theory , volume =

    Duggan, John and Fey, Mark , title =. International Journal of Game Theory , volume =

  17. [17]

    arXiv preprint arXiv:2603.08752 , year =

    Mukherjee, Sumit , title =. arXiv preprint arXiv:2603.08752 , year =

  18. [18]

    American Political Science Review , year =

    Laver, Michael , title =. American Political Science Review , year =