pith. sign in

arxiv: 2602.06342 · v2 · pith:WHKF6KUCnew · submitted 2026-02-06 · ⚛️ physics.soc-ph · cond-mat.stat-mech

Mean-Field Theory for Heider Balance under Heterogeneous Social Temperatures

Zhen Li , Yuki Izumida This is my paper

Pith reviewed 2026-05-16 06:57 UTC · model grok-4.3

classification ⚛️ physics.soc-ph cond-mat.stat-mech
keywords Heider balanceheterogeneous temperaturesmean-field theoryphase transitionssocial networksstructural balanceopinion dynamicsdistribution dependence
0
0 comments X

The pith

A mean-field model shows that the full distribution of social temperatures across links determines the transition to polarized states in Heider balance.

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

This paper generalizes the Heider balance model to allow each relationship its own social temperature, capturing varying levels of stability and change in social ties. In a mean-field treatment on the complete graph, a self-consistency condition is derived that depends on the entire distribution of these temperatures to locate the stable collective states. The analysis reveals that light-tailed versus heavy-tailed forms of the inverse-temperature distribution produce distinct phase diagrams for polarization. The uniform temperature case sets the lowest critical threshold for the average inverse temperature at which polarization emerges. Simulations on the complete graph validate these distribution-dependent predictions.

Core claim

Within the mean-field formulation of the Heider balance model with link-specific social temperatures, the collective opinion state obeys a self-consistency condition that incorporates the full distribution of inverse temperatures. This condition identifies the criteria for the transition between polarized and non-polarized configurations, with the functional form of the distribution leading to qualitatively different phase behaviors. The homogeneous-temperature limit serves as a universal lower bound on the critical mean inverse temperature required for polarization.

What carries the argument

The distribution-dependent self-consistency equation for the mean opinion, which encodes how heterogeneity in inverse temperatures controls the macroscopic polarization.

If this is right

  • Polarization occurs only when the mean inverse temperature exceeds a threshold set by the specific distribution shape.
  • Heavy-tailed distributions can lead to different transition points compared to light-tailed ones.
  • The critical value is always at least as high as in the uniform temperature case.
  • The framework extends naturally to other distributions chosen by the modeler.

Where Pith is reading between the lines

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

  • Real-world social data on relation volatility distributions could be used to forecast overall network balance.
  • Incorporating network sparsity might show how local heterogeneity amplifies or dampens the mean-field effects.
  • Adding time-varying temperatures could reveal dynamic transitions not captured in the static distribution case.

Load-bearing premise

The mean-field approximation remains valid on the complete graph when temperatures are independently drawn once from any fixed distribution.

What would settle it

Run Monte Carlo simulations of the stochastic Heider dynamics on a large complete graph with a chosen heavy-tailed inverse-temperature distribution and measure whether the observed onset of polarization occurs at the mean-field predicted critical mean inverse temperature.

Figures

Figures reproduced from arXiv: 2602.06342 by Yuki Izumida, Zhen Li.

Figure 1
Figure 1. Figure 1: FIG. 1: (a) Schematics of a complete graph for Heider [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: shows the mean-field solutions of Eq. (8) as the function of µ −1 for σ = 1 under the gamma distribu￾tion (16). The system has a pair of stable and unstable polarized solutions ⟨x⟩ ̸= 0 below a critical point, while the trivial non-polarized solution ⟨x⟩ = 0 always exists. In accordance with the mean-field prediction for this het￾erogeneous system, the simulation results of (2) show the discontinuous phase… view at source ↗
Figure 3
Figure 3. Figure 3: shows the phase diagram with respect to (σ, µ−1 ) under the gamma distribution (16). The crit￾ical curve separating the three fixed-points area and the only one fixed-point area is located between the upper and lower bounds in Eq. (11). The upper bound line, σ 2 = µ 2 100 r 3125 + 6144µ2 5 − 75! , (17) is calculated from the right-hand side of Eq. (11) by con￾sidering γ 3 = α(α + 1)(α + 2)θ 3 for the gamma… view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4: Mean-field solutions of [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6: Phase diagram with respect to ( [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗
read the original abstract

Heider balance theory provides a fundamental framework for understanding the formation of friendly and hostile relations in social networks. Existing stochastic formulations typically assume a uniform social temperature, implying that all interpersonal relations fluctuate with the same intensity. However, studies show that social interactions are highly heterogeneous, with broad variability in stability, volatility, and susceptibility to change. In this work, we introduce a generalized Heider balance model on a complete graph in which each link is assigned its own social temperature. Within a mean-field formulation, we derive a distribution-dependent self-consistency condition for the collective opinion state and identify the criteria governing the transition between polarized and non-polarized configurations. This framework reveals how the entire distribution of interaction heterogeneity shapes the macroscopic behavior of the system. We show that the functional form of the inverse-temperature distribution, in particular whether it is light-tailed or heavy-tailed, leads to qualitatively distinct phase diagrams. We also establish universal bounds for the critical transition, where the homogeneous-temperature limit provides a universal lower bound for the critical mean of an inverse-temperature distribution governing the transition. Numerical simulations confirm the theoretical predictions and highlight the nontrivial effects introduced by heterogeneity. Our results provide a unified route to understanding structural balance in realistic social systems and lay the groundwork for extensions incorporating fluctuations beyond mean field, external fields, and network topologies beyond the complete graph.

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 / 2 minor

Summary. The manuscript develops a mean-field theory for the Heider balance model on complete graphs with heterogeneous link temperatures drawn from an arbitrary distribution P(β). It derives a distribution-dependent self-consistency condition for the collective opinion state (order parameter), identifies criteria for transitions between polarized and non-polarized configurations, shows that light-tailed versus heavy-tailed forms of P(β) produce qualitatively distinct phase diagrams, establishes universal bounds on the critical mean inverse temperature (with the homogeneous-temperature limit as a lower bound), and confirms the predictions via numerical simulations.

Significance. If the mean-field closure holds, the work provides a unified analytic route to incorporating realistic heterogeneity in interaction stability into structural balance theory. The distribution-dependent self-consistency condition and the universal bound on the critical mean are potentially useful results that could guide extensions to fluctuating dynamics, external fields, and non-complete topologies.

major comments (2)
  1. [mean-field derivation section] Self-consistency derivation (section containing the mean-field closure, likely §3): the replacement of heterogeneous β_ij by an integral over P(β) inside a single global order-parameter equation assumes that every link contributes statistically identically and that fluctuations around the mean opinion remain small. For heavy-tailed P(β) where the second moment diverges, rare low-temperature links can locally pin balanced configurations and bias the global state; the manuscript supplies no fluctuation analysis or error bound that remains valid in this regime.
  2. [phase diagram section] Phase-diagram claims for heavy-tailed distributions (section presenting the light- vs. heavy-tailed results): the reported qualitative distinction between light- and heavy-tailed phase diagrams rests on the mean-field approximation remaining accurate. Without an explicit test (e.g., comparison of the self-consistency solution against Monte Carlo results on finite complete graphs for a Pareto distribution with diverging variance), the central claim that the entire distribution shapes macroscopic behavior cannot be verified.
minor comments (2)
  1. [abstract and introduction] Notation: the abstract and main text alternate between 'social temperature' and 'inverse-temperature distribution' without an early explicit statement of the convention β = 1/T; a single sentence defining the mapping would remove ambiguity.
  2. [numerical results section] Figure clarity: the numerical confirmation plots should include error bars or shaded regions indicating sample-to-sample variability across independent draws of the temperature distribution, especially for heavy-tailed cases.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the thoughtful review and valuable comments on our manuscript. We address the major concerns point by point below, providing clarifications and indicating revisions where appropriate.

read point-by-point responses

  1. Referee: [mean-field derivation section] Self-consistency derivation (section containing the mean-field closure, likely §3): the replacement of heterogeneous β_ij by an integral over P(β) inside a single global order-parameter equation assumes that every link contributes statistically identically and that fluctuations around the mean opinion remain small. For heavy-tailed P(β) where the second moment diverges, rare low-temperature links can locally pin balanced configurations and bias the global state; the manuscript supplies no fluctuation analysis or error bound that remains valid in this regime.

    Authors: The mean-field closure is obtained by replacing the heterogeneous interactions with their average over P(β) under the assumption that the global order parameter captures the dominant behavior. While we agree that fluctuations could be significant for heavy-tailed distributions, our derivation is exact in the thermodynamic limit for the complete graph, and the numerical simulations presented in the manuscript show quantitative agreement with the self-consistency equation even for heavy-tailed P(β). We have added a paragraph in the discussion section acknowledging the potential role of rare events and the absence of a rigorous fluctuation analysis, which is left for future work. revision: partial

  2. Referee: [phase diagram section] Phase-diagram claims for heavy-tailed distributions (section presenting the light- vs. heavy-tailed results): the reported qualitative distinction between light- and heavy-tailed phase diagrams rests on the mean-field approximation remaining accurate. Without an explicit test (e.g., comparison of the self-consistency solution against Monte Carlo results on finite complete graphs for a Pareto distribution with diverging variance), the central claim that the entire distribution shapes macroscopic behavior cannot be verified.

    Authors: We have conducted additional Monte Carlo simulations specifically for Pareto distributions with shape parameters leading to diverging variance (e.g., α=1.5). These simulations on finite complete graphs (N=100 to 500) confirm that the mean-field predictions for the critical mean inverse temperature and the phase boundaries hold with good accuracy, supporting the qualitative distinction. The new results are incorporated into the revised manuscript as an additional figure and accompanying text. revision: yes

standing simulated objections not resolved
  • A detailed fluctuation analysis or derivation of error bounds for the mean-field approximation in regimes where the variance of P(β) diverges.

Circularity Check

0 steps flagged

No significant circularity; self-consistency derived from model equations

full rationale

The derivation begins from the stochastic Heider balance dynamics on the complete graph with link-specific inverse temperatures β_ij drawn from a chosen distribution P(β). The mean-field closure replaces the local fields with a global order parameter m and obtains a self-consistency equation by averaging the hyperbolic-tangent response over P(β). The critical mean β_c is located by linearizing this integral equation around m=0 and finding the point at which a nonzero solution appears; this bifurcation condition is an algebraic consequence of the assumed dynamics and the mean-field ansatz, not a redefinition or fit of the target quantity. No load-bearing self-citations, no fitted parameters relabeled as predictions, and no ansatz smuggled via prior work are required for the central result. The framework therefore remains self-contained against the stated model equations.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The model rests on a chosen distribution of inverse temperatures (a modeling choice) and the mean-field closure on the complete graph. No new particles or forces are postulated.

free parameters (1)
  • inverse-temperature distribution
    The functional form (light- or heavy-tailed) and any parameters inside it are chosen by the modeler to represent heterogeneity; the critical mean is then computed from that choice.
axioms (1)
  • domain assumption Mean-field approximation is valid for the complete graph with independent link temperatures
    Invoked to close the equations for the collective opinion state.

pith-pipeline@v0.9.0 · 5533 in / 1355 out tokens · 50458 ms · 2026-05-16T06:57:46.402983+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

54 extracted references · 54 canonical work pages · 2 internal anchors

  1. [1]

    Friends of my friends are my friends

    work page

  2. [2]

    Enemies of my friends are my enemies

    work page

  3. [3]

    Friends of my enemies are my enemies

    work page

  4. [4]

    Mean-Field Theory for Heider Balance under Heterogeneous Social Temperatures

    Enemies of my enemies are my friends. These principles can be described by triadic balance on social networks [6]. Since the seminal works of Heider [1], and Cartwright and Harary [3], triadic balance has become a cornerstone concept in social groups [7–9]. The triadic balance may be considered on signed networks with plus (friend) or minus (enemy) sign b...

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  5. [5]

    1” to “2

    Here,C ∗ ≈1.716 is the critical value reached for a delta distribution [11]. The derivation of the bounds (11) is provided in Ap- pendix B The lower bound indicates that the homogeneous tem- perature case achieves the highest critical temperature for the polarized state. This can be understood in terms of the role of low temperature links in the heterogen...

    work page

  6. [6]

    Appendix C: Proof of the asymptotic behavior for the critical mean The proof of Eq

    The range ofγis required due toγ≥µ derived above, which comes fromγ≥ 5 4 4 3 γ3 1 5 . Appendix C: Proof of the asymptotic behavior for the critical mean The proof of Eq. (12) can be considered in the following way. We first consider the case where the distribution ρ(β) vanishes asλ→+∞for allβ >0, which means that for a distribution familyρ λ(β) parameteri...

    work page

  7. [7]

    Heider, J

    F. Heider, J. Psychol.21, 107 (1946)

    work page 1946

  8. [8]

    T. C. Blum, Am. J. Sociol.91, 511 (1985)

    work page 1985

  9. [9]

    Cartwright and F

    D. Cartwright and F. Harary, Psychol. Rev.63, 277 (1956)

    work page 1956

  10. [10]

    M. S. Berk and S. M. Andersen, J. Pers. Soc. Psychol. 79, 546 (2000)

    work page 2000

  11. [11]

    Malarz and K

    K. Malarz and K. Ku lakowski, Physica A567, 125640 (2021)

    work page 2021

  12. [12]

    Antal, P

    T. Antal, P. L. Krapivsky, and S. Redner, Phys. Rev. E 72, 036121 (2005)

    work page 2005

  13. [13]

    Altafini, PloS One7, e38135 (2012)

    C. Altafini, PloS One7, e38135 (2012)

    work page 2012

  14. [14]

    J. Lan, Y. Huo, Z. Cai, C.-S. Wong, Z. Chen, and W. Lam, J. Occup. Organ. Psychol.93, 654 (2020)

    work page 2020

  15. [15]

    Corbetta and K

    R. Corbetta and K. A. Grant, Confl. Manag. Peace Sci. 29, 314 (2012)

    work page 2012

  16. [16]

    Zheng, D

    X. Zheng, D. Zeng, and F.-Y. Wang, Inf. Syst. Front.17, 1077 (2015)

    work page 2015

  17. [17]

    Malarz and J

    K. Malarz and J. A. Ho lyst, Phys. Rev. E106, 064139 (2022)

    work page 2022

  18. [18]

    Rabbani, A

    F. Rabbani, A. H. Shirazi, and G. Jafari, Phys. Rev. E 99, 062302 (2019)

    work page 2019

  19. [19]

    Bukina and D

    K. Bukina and D. L. Shepelyansky, Opinion forma- tion at ising social networks (2025), arXiv:2511.12786 [physics.soc-ph]

    work page arXiv 2025

  20. [20]

    Relationship between Heider links and Ising spins

    Z. Burda, M. Wo loszyn, K. Malarz, and K. Ku lakowski, On the relationship between heider links and ising spins (2025), arXiv:2512.02644 [cond-mat.stat-mech]

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  21. [21]

    Burda, M

    Z. Burda, M. Wo loszyn, K. Malarz, and K. Ku lakowski, Heider balance of a square lattice in an external field (2025), arXiv:2512.00567 [cond-mat.stat-mech]

    work page arXiv 2025

  22. [22]

    H. Goto, M. Shiraishi, and H. Nishimori, Phys. Rev. Lett. 133, 127402 (2024)

    work page 2024

  23. [23]

    A. M. Belaza, K. Hoefman, J. Ryckebusch, A. Bram- son, M. van den Heuvel, and K. Schoors, PLoS One12, e0183696 (2017)

    work page 2017

  24. [24]

    H. Du, X. He, J. Wang, and M. W. Feldman, Physica A 503, 780 (2018)

    work page 2018

  25. [25]

    A. M. Belaza, J. Ryckebusch, A. Bramson, C. Casert, K. Hoefman, K. Schoors, M. van den Heuvel, and B. Van- dermarliere, Physica A518, 270 (2019)

    work page 2019

  26. [26]

    Kirkley, G

    A. Kirkley, G. T. Cantwell, and M. E. Newman, Phys. Rev. E99, 012320 (2019)

    work page 2019

  27. [27]

    Bagherikalhor, A

    M. Bagherikalhor, A. Kargaran, A. Shirazi, and G. Jafari, Phys. Rev. E103, 032305 (2021)

    work page 2021

  28. [28]

    Babaee, M

    Z. Babaee, M. Bagherikalhor, L. Elyasizad, M. D. Niry, and G. R. Jafari, Phys. Rev. E105, 044307 (2022)

    work page 2022

  29. [29]

    Hao and I

    B. Hao and I. A. Kov´ acs, Sci. Adv.10, eadj0104 (2024)

    work page 2024

  30. [30]

    M. J. Krawczyk, M. Wo loszyn, P. Gronek, K. Ku lakowski, and J. Mucha, Sci. Rep.9, 11202 (2019)

    work page 2019

  31. [31]

    Y. Zhao, L. Zhang, M. Tang, and G. Kou, Comput. Oper. Res.74, 205 (2016)

    work page 2016

  32. [32]

    Stern and G

    S. Stern and G. Livan, R. Soc. Open Sci.8, 201943 (2021)

    work page 2021

  33. [33]

    Lambiotte, M

    R. Lambiotte, M. Rosvall, and I. Scholtes, Nat. Phys.15, 313 (2019)

    work page 2019

  34. [34]

    Battiston, E

    F. Battiston, E. Amico, A. Barrat, G. Bianconi, G. Fer- raz de Arruda, B. Franceschiello, I. Iacopini, S. K´ efi, V. Latora, Y. Moreno, M. M. Murray, T. P. Peixoto, F. Vaccarino, and G. Petri, Nat. Phys.17, 1093 (2021)

    work page 2021

  35. [35]

    Schawe and L

    H. Schawe and L. Hern´ andez, Commun. Phys.5, 32 (2022)

    work page 2022

  36. [36]

    Xu and Z

    W. Xu and Z. Zhang, IEEE Trans. Comput. Soc. Syst 11, 3627 (2023)

    work page 2023

  37. [37]

    Zhang, M

    Y. Zhang, M. Lucas, and F. Battiston, Nat. Commun. 14, 1605 (2023)

    work page 2023

  38. [38]

    Mohandas, K

    K. Mohandas, K. Suchecki, and J. A. Ho lyst, Phys. Rev. E111, 024310 (2025)

    work page 2025

  39. [39]

    Varshney, A

    D. Varshney, A. K. Mishra, P. K. Joshi, and D. Roy, Food Policy112, 102360 (2022)

    work page 2022

  40. [40]

    Dakin and T

    R. Dakin and T. B. Ryder, Proc. Natl. Acad. Sci. U.S.A. 117, 2993 (2020)

    work page 2020

  41. [41]

    Z. Wu, Y. Mu, and Z. Wang, Plos One20, e0332474 (2025)

    work page 2025

  42. [42]

    D. W. Johnson, R. T. Johnson, and G. Maruyama, Rev. Educ. Res.53, 5 (1983)

    work page 1983

  43. [43]

    Eom and H.-H

    Y.-H. Eom and H.-H. Jo, Sci. Rep.5, 09752 (2015)

    work page 2015

  44. [44]

    Enders, H

    A. Enders, H. Hungenberg, H.-P. Denker, and S. Mauch, Eur. Manag. J26, 199 (2008)

    work page 2008

  45. [45]

    J. Zhao, X. Wang, C. Gu, and Y. Qin, Dyn. Games Appl. 11, 612 (2021)

    work page 2021

  46. [46]

    Ferreri, P

    L. Ferreri, P. Bajardi, M. D. L. Giacobini, S. Perazzo, E. Venturino,et al., Phys. Rev. E90, 012812 (2014)

    work page 2014

  47. [47]

    Alexandrov and G

    A. Alexandrov and G. S. Medvedev, Phase transi- tions in the ising model on random graphs (2025), arXiv:2511.10838 [math-ph]

    work page arXiv 2025

  48. [48]

    Robiglio, L

    T. Robiglio, L. Di Gaetano, A. Altieri, G. Petri, and F. Battiston, Phys. Rev. E112, L022301 (2025)

    work page 2025

  49. [49]

    R. J. Glauber, J. Math. Phys.4, 294 (1963)

    work page 1963

  50. [50]

    T. R. Kirkpatrick and D. Thirumalai, Phys. Rev. B36, 5388 (1987)

    work page 1987

  51. [51]

    M´ ezard, G

    M. M´ ezard, G. Parisi, and M. A. Virasoro,Spin Glass Theory and Beyond: An Introduction to the Replica Method and Its Applications(World Scientific Publish- ing Company, 1987)

    work page 1987

  52. [52]

    Nishimori,Statistical Physics of Spin Glasses and Information Processing: An Introduction(Clarendon Press, 2001)

    H. Nishimori,Statistical Physics of Spin Glasses and Information Processing: An Introduction(Clarendon Press, 2001)

    work page 2001

  53. [53]

    Umantsev,Field theoretic method in phase transfor- mations, Vol

    A. Umantsev,Field theoretic method in phase transfor- mations, Vol. 840 (Springer, 2012)

    work page 2012

  54. [54]

    S. S. Dragomir and N. M. Ionescu, Rev. Anal. Num´ er. Th´ eor. Approx.23, 71 (1994)

    work page 1994