Pith. sign in

REVIEW 4 minor 19 references

Scoring rules are least-squares estimators of score vectors, which is why they coincide with cosine similarity rankings.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-14 13:56 UTC pith:WVN6B6JO

load-bearing objection Clean, elementary re-proof of Kawada’s equivalence that makes the geometry obvious; incremental but solid note.

arxiv 2607.10145 v1 pith:WVN6B6JO submitted 2026-07-11 econ.TH

Scoring Rules as Least-Squares Estimators

Satoru Fujishige , Satoshi Nakada This is my paper
classification econ.TH
keywords scoring rulesBorda ruleleast-squarescosine similaritysocial choiceoptimizationgeometric interpretation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This note re-proves that every scoring rule in voting theory is identical to the corresponding cosine-similarity rule. The authors show that the arithmetic mean of voters' score vectors is the unique point that minimizes the sum of squared Euclidean distances to those vectors. Because every score vector has the same length, that same mean also maximizes the sum of cosine similarities. Ranking alternatives by the coordinates of the mean therefore yields exactly the ranking produced by ordinary score aggregation. The geometric picture makes the earlier equivalence transparent: both procedures are different faces of the same least-squares estimator.

Core claim

For any fixed score vector θ, the social ranking obtained by summing the individual score vectors is identical to the ranking obtained by maximizing the sum of cosine similarities with those vectors. The common optimizer is the arithmetic mean of the score vectors, which is uniquely characterized as the least-squares estimator of those vectors.

What carries the argument

The elementary least-squares fact that the arithmetic mean uniquely minimizes the sum of squared Euclidean distances; once all score vectors are known to have identical Euclidean norm, this fact immediately converts the least-squares problem into the cosine-similarity maximization problem.

Load-bearing premise

Every voter's score vector has exactly the same Euclidean length, which follows only because the underlying score schedule θ is the same for every voter.

What would settle it

Construct a profile of score vectors that do not all share the same Euclidean norm and check whether the least-squares mean and the cosine-similarity maximizer still induce the same ranking of alternatives; any divergence would break the claimed equivalence.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

0 major / 4 minor

Summary. This note re-proves Kawada (2018)’s equivalence between scoring rules and cosine-similarity rules. After recalling that the arithmetic mean uniquely minimizes the sum of squared Euclidean distances (Proposition 1), the authors observe that every individual score vector s_i^θ has the same Euclidean norm R = (∑_k θ_k^{2})^{1/2}. Consequently the least-squares problem is equivalent, on any sphere of positive radius, to maximizing the sum of cosines of the angles between the score vectors and the decision variable. The common optimizer is therefore the mean score vector, which induces the same ranking as the ordinary scoring rule (Theorem 1).

Significance. The contribution is a short, transparent geometric re-proof of a known equivalence. By identifying scoring rules as least-squares estimators of the individual score vectors, the note supplies a clean geometric rationale for why cosine similarity necessarily recovers the same social ranking. The argument is elementary, fully self-contained, and free of circularity or free parameters. While the result itself is not new, the least-squares perspective is pedagogically useful and may open the door to analogous geometric treatments of weighted scoring rules or incomplete rankings, as the authors themselves suggest.

minor comments (4)
  1. The date on the title page is given as July 14, 2026; this is almost certainly a typographical error and should be corrected before publication.
  2. In the definition of the ranking functions (Section 3.1) the authors write P ⊊ R for the set of strict rankings; the proper symbol is the proper-subset symbol ⊂, or simply ⊆ if equality is allowed.
  3. The phrase “its subset P ⊊ R denote the set of all strict rankings” is grammatically incomplete; insert “let” or rephrase.
  4. A brief parenthetical remark that the same argument applies verbatim to any collection of vectors of equal length (not merely score vectors) would make the geometric content even clearer.

Circularity Check

0 steps flagged

No circularity: elementary least-squares identity plus constant-norm fact yields Kawada's equivalence without self-reference or fitted parameters.

full rationale

The paper restates Kawada (2018) Theorem 1 and supplies an alternative elementary proof. Proposition 1 is the classical fact that the arithmetic mean uniquely minimizes sum of squared Euclidean distances; it is derived in closed form from expanding Phi(x) and does not presuppose the target ranking equivalence. The only additional algebraic step is that every individual score vector s_i^theta has identical Euclidean norm R = (sum_k theta_k^2)^{1/2}, which follows immediately from the definition of a scoring rule (the same score vector theta is applied to every ranking). Consequently the least-squares objective restricted to any sphere of positive radius is a positive multiple of the sum of cosines, so the two optimizers coincide and induce the same ranking. No parameter is fitted to data, no uniqueness theorem is imported from the authors' prior work, and the external citation to Kawada is used only as the statement being re-proved, not as a load-bearing premise. The derivation is therefore self-contained and non-circular.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 0 invented entities

The note relies only on standard Euclidean geometry and the classical definition of scoring rules; no free parameters are fitted and no new entities are postulated. The sole domain assumptions are the usual ones of social-choice theory (strict rankings, fixed score vector θ).

axioms (3)
  • standard math The arithmetic mean uniquely minimizes the sum of squared Euclidean distances (Proposition 1).
    Invoked as the starting point of the proof of Theorem 1; classical fact from least-squares theory.
  • domain assumption Every individual score vector s_i^θ is a rearrangement of the same fixed vector θ, hence all have identical Euclidean norm.
    Used in the proof of Theorem 1 to pull the constant R out of the inner-product sum and equate it with the cosine objective.
  • domain assumption Voters submit strict total rankings and the social ranking is the weak order induced by the coordinates of an optimal vector.
    Standard social-choice setting stated in Section 3; needed for the ranking comparison (2) and (4) to be well-defined.

pith-pipeline@v1.1.0-grok45 · 9246 in / 1863 out tokens · 30354 ms · 2026-07-14T13:56:32.293217+00:00 · methodology

0 comments
read the original abstract

Kawada (2018) proved that every scoring rule is equivalent to the corresponding cosine similarity rule. The original proof relies on a direct analysis of the cosine similarity optimization problem. In this note, we present an alternative, simpler proof based on a basic least-squares characterization. Our argument shows that the arithmetic mean of the score vectors is the unique minimizer of the total squared Euclidean distance and that the cosine similarity formulation is an immediate consequence of this optimization property. This result provides a transparent geometric interpretation of scoring rules and clarifies why the cosine similarity rule necessarily coincides with the corresponding scoring rule.

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

19 extracted references · 1 linked inside Pith

  1. [1]

    Social Choice and Welfare , volume =

    Kawada, Yoko , title =. Social Choice and Welfare , volume =. 2018 , doi =

    2018

  2. [2]

    Peyton , title =

    Young, H. Peyton , title =. Journal of Economic Theory , volume =. 1974 , doi =

    1974

  3. [3]

    , title =

    Saari, Donald G. , title =. Social Choice and Welfare , volume =. 1990 , doi =

    1990

  4. [4]

    , title =

    Saari, Donald G. , title =. 1995 , series =

    1995

  5. [5]

    Public Choice , volume =

    Nitzan, Shmuel and Rubinstein, Ariel , title =. Public Choice , volume =

  6. [6]

    , title =

    Fishburn, Peter C. , title =

  7. [7]

    Peyton Young , title =

    H. Peyton Young , title =. American Political Science Review , volume =

  8. [8]

    Kemeny , title =

    John G. Kemeny , title =. Daedalus , volume =. 1959 , url =

    1959

  9. [9]

    1997 , month = oct, url =

    Markus Schulze , title =. 1997 , month = oct, url =

    1997

  10. [11]

    Fishburn, P. C. (1973): The Theory of Social Choice, Princeton: Princeton University Press

    1973

  11. [12]

    Goto, and S

    Fujishige, S., L. Goto, and S. Nakada (2026): A New Method for Finding the Schulze Winner Set, arXiv preprint arXiv:2606.02213

    Pith/arXiv arXiv 2026

  12. [13]

    (2018): Cosine Similarity and the Borda Rule, Social Choice and Welfare, 51, 1--11

    Kawada, Y. (2018): Cosine Similarity and the Borda Rule, Social Choice and Welfare, 51, 1--11

    2018

  13. [14]

    Kemeny, J. G. (1959): Mathematics without Numbers, Daedalus, 88, 577--591

    1959

  14. [15]

    Nitzan, S. and A. Rubinstein (1981): A Further Characterization of Borda Ranking Method, Public Choice, 36, 153--158

    1981

  15. [16]

    Saari, D. G. (1990): The Borda Dictionary, Social Choice and Welfare, 7, 279--317

    1990

  16. [17]

    --- -.1pt --- -.1pt --- (1995): Basic Geometry of Voting, Studies in Economic Theory, Berlin: Springer

    1995

  17. [18]

    (1997): Condorcet Sub-Cycle Rule, Election Methods Mailing List, first public description of the Schulze method

    Schulze, M. (1997): Condorcet Sub-Cycle Rule, Election Methods Mailing List, first public description of the Schulze method

    1997

  18. [19]

    Young, H. P. (1974): An Axiomatization of Borda's Rule, Journal of Economic Theory, 9, 43--52

    1974

  19. [20]

    --- -.1pt --- -.1pt --- (1988): Condorcet's Theory of Voting, American Political Science Review, 82, 1231--1244

    1988