pith. sign in

arxiv: 2605.20979 · v1 · pith:MJ3F2D4Onew · submitted 2026-05-20 · ⚛️ physics.soc-ph · cond-mat.stat-mech

Equilibrium and dynamics of a three-state opinion model on a network of networks

Irene Ferri , Albert D\'iaz-Guilera , Hiroki Sayama This is my paper

Pith reviewed 2026-05-21 02:09 UTC · model grok-4.3

classification ⚛️ physics.soc-ph cond-mat.stat-mech
keywords opinion dynamicsnetwork of networksthree-state modelinternal belief topologypolarized consensuscritical temperatureneutrality parameterMonte Carlo simulations
0
0 comments X

The pith

The internal organization of beliefs within each agent determines how much social agitation is needed to destabilize a polarized group consensus.

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

This paper studies a three-state opinion model on a network of networks where each agent has an internal graph of beliefs that can each be in one of two polarized states or neutral. It explores how different internal topologies affect the stability of collective polarized opinions when agents interact through an external social graph, with temperature representing both external agitation and tolerance for internal inconsistencies. Using a fully connected external graph and internal structures given by chains, cliques, and star-like forms with a central core belief, the work shows that the critical temperature at which polarized consensus breaks increases with added beliefs only for star-like agents and saturates for the other topologies. Binary mixtures of agents with different internal topologies are also examined, with the dominant type depending on the neutrality parameter. A sympathetic reader would care because real individuals hold multiple linked beliefs whose arrangement may shape how groups respond to noise or persuasion.

Core claim

Each agent holds an internal belief graph where beliefs take polarized or neutral values, with a neutrality parameter controlling neutral conviction and temperature accounting for external agitation and internal dissonance. On a fully connected external social graph with internal topologies of one-dimensional chains, cliques, and star-like structures featuring a central core belief, the polarized consensus destabilizes at a critical temperature that increases with added beliefs for star-like agents but saturates for ring- and clique-like topologies. In equal-proportion binary mixtures of different topologies, the dominant influence on collective behavior depends on the neutrality parameter.

What carries the argument

The network-of-networks structure, where a fully connected external social graph links agents each carrying an internal belief topology (one-dimensional chains, cliques, or stars with central core), with temperature controlling both inter-agent interactions and intra-agent consistency.

If this is right

  • Polarized consensus becomes more stable against rising temperature when agents have star-like internal belief structures with more beliefs attached.
  • For ring and clique internal topologies the critical temperature saturates and stops increasing once a certain number of beliefs is reached.
  • In equal mixtures of agents with different internal topologies, which type dominates the collective outcome depends on the value of the neutrality parameter.
  • The model predicts regime-dependent interplay between agents of differing internal organizations.

Where Pith is reading between the lines

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

  • Real populations containing many individuals with core beliefs linked to multiple peripherals may sustain consensus under higher levels of social noise than populations with densely connected belief sets.
  • The observed saturation effect implies there may be an effective limit on how much internal complexity improves stability for chain or clique topologies.
  • Extending the external graph to include community structure or sparse connections could show whether the topology-specific temperature shifts survive outside the complete-graph case.

Load-bearing premise

The external social graph is fully connected and internal belief structures are restricted to simple topologies like one-dimensional chains, cliques, and stars with a central core belief.

What would settle it

A Monte Carlo run on the same model but with a non-fully-connected external graph in which the critical temperature fails to rise with added beliefs for star-like agents would falsify the reported dependence.

Figures

Figures reproduced from arXiv: 2605.20979 by Albert D\'iaz-Guilera, Hiroki Sayama, Irene Ferri.

Figure 1
Figure 1. Figure 1: FIG. 1. Representation of four agents with a 4-node ring [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Illustration of the three network topologies used to model belief systems with a number of internal beliefs [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Contribution of a pair of interacting agents to the [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Magnetization curves for a particular belief as a function of temperature for each internal topology and different [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Critical temperature versus number of beliefs when [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. Magnetization curves for a particular belief as a function of temperature for each internal topology, and a different [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. Example of two star-like agents with three in- ternal [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9. Distribution of internal parameters per agent for [PITH_FULL_IMAGE:figures/full_fig_p010_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10. Distribution of the absolute values of internal magnetization (first column) and the number of internal neutral beliefs [PITH_FULL_IMAGE:figures/full_fig_p012_10.png] view at source ↗
read the original abstract

Opinion formation models typically represent each individual as a single variable. However, in practice each individual holds interconnected beliefs whose internal organization may influence collective outcomes. To explore this dependence, we study a three-state opinion model on a network of networks in which each agent has an internal belief graph and interacts with other agents through an external social graph. Each belief can take two opposite polarized states or a neutral one and a neutrality parameter tunes the relative conviction of the neutral stance. We incorporate temperature into the model to account for external social agitation and for the tolerance of internal cognitive dissonance. We explore the stationary state and dynamics of the model using analytical approaches and Monte Carlo simulations on a fully connected external social graph, with internal belief topologies given by one-dimensional chains, cliques, and star-like structures, where there is a central core belief to which all other beliefs are connected. We find that the critical temperature at which the polarized consensus destabilizes increases with the addition of more beliefs to star-like agents but saturates in the case of ring- and clique-like internal topologies. We also consider binary mixtures of agents with different internal topologies in equal proportions, showing that the interplay between agents is regime-dependent, with the dominant topology depending on the value of the neutrality parameter.

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

1 major / 2 minor

Summary. The manuscript introduces a three-state opinion model on a network of networks in which each agent is equipped with an internal belief graph. Each belief takes one of two polarized states or a neutral state, with a neutrality parameter controlling the weight of the neutral stance. Agents interact through a fully connected external social graph, and temperature is introduced to model social agitation and tolerance to internal dissonance. Analytical methods and Monte Carlo simulations are used to examine stationary states and dynamics for internal topologies consisting of one-dimensional chains, cliques, and star-like structures (with a central core belief). The central result is that the critical temperature at which polarized consensus destabilizes increases with the number of beliefs for star-like agents but saturates for ring- and clique-like internal topologies. Binary mixtures of agents with different internal topologies are also studied, revealing regime-dependent dominance governed by the neutrality parameter.

Significance. If the reported topology dependence survives normalization checks, the work provides a useful extension of opinion dynamics models by showing how the internal organization of an agent's beliefs can modulate collective stability against temperature. The combination of analytical treatment and Monte Carlo simulations on concrete topologies, together with the analysis of mixed populations, constitutes a clear strength and opens avenues for studying cognitive structure effects in sociophysics.

major comments (1)
  1. [§2] §2 (model definition): The internal Hamiltonian is written as a direct sum of interactions over the edges of the belief graph with no rescaling by the number of beliefs or the number of edges. Consequently, clique and ring topologies accumulate interaction energy quadratically or linearly with added beliefs, while star topologies accumulate it linearly; this difference in total coupling strength at fixed temperature could produce the reported saturation versus monotonic increase without any topological mechanism. The central claim comparing critical temperatures across topologies therefore requires either an explicit normalization (e.g., division by edge count) or a supplementary check that the qualitative behavior is invariant under such rescaling.
minor comments (2)
  1. [Abstract] The abstract refers to 'ring- and clique-like internal topologies' while the model section specifies 'one-dimensional chains, cliques, and star-like structures'; a brief clarification of whether rings are distinct from chains would improve consistency.
  2. [Figures] Figure captions and legends should explicitly state the number of Monte Carlo realizations and whether error bars represent standard deviations or standard errors, especially for the critical-temperature curves.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their detailed and constructive report. The concern regarding the normalization of the internal Hamiltonian is well taken, and we address it directly below by clarifying the modeling choice and reporting an explicit check that preserves the central claims.

read point-by-point responses

  1. Referee: [§2] §2 (model definition): The internal Hamiltonian is written as a direct sum of interactions over the edges of the belief graph with no rescaling by the number of beliefs or the number of edges. Consequently, clique and ring topologies accumulate interaction energy quadratically or linearly with added beliefs, while star topologies accumulate it linearly; this difference in total coupling strength at fixed temperature could produce the reported saturation versus monotonic increase without any topological mechanism. The central claim comparing critical temperatures across topologies therefore requires either an explicit normalization (e.g., division by edge count) or a supplementary check that the qualitative behavior is invariant under such rescaling.

    Authors: We agree that the unnormalized sum over edges leads to different total internal coupling strengths across topologies when the number of beliefs increases. Our modeling choice treats each belief–belief interaction as an independent unit-strength constraint, so that adding beliefs naturally augments the total internal field experienced by the agent; this is intentional and reflects the interpretation that more beliefs impose more cognitive constraints. Nevertheless, to isolate topological effects from the overall energy scale, we have repeated the Monte Carlo simulations after rescaling the internal Hamiltonian by the number of edges in each belief graph. The qualitative results remain unchanged: the critical temperature at which polarized consensus destabilizes continues to rise monotonically with the number of beliefs for star-like agents, while it saturates for both ring and clique topologies. We will add a brief discussion of this normalization check together with the corresponding supplementary figure in the revised §2 and results sections. revision: yes

Circularity Check

0 steps flagged

No significant circularity; results follow from explicit model definitions and simulations

full rationale

The paper defines a three-state opinion model on a network of networks with explicit internal belief topologies (chains, cliques, stars) and an external fully connected social graph, plus parameters for neutrality and temperature. Stationary states and critical temperatures are obtained from analytical mean-field approaches and Monte Carlo simulations on these fixed structures. No derivation step reduces a claimed result to an input by construction, no fitted quantity is relabeled as a prediction, and no load-bearing premise rests on self-citation chains or imported uniqueness theorems. The reported saturation versus monotonic increase in critical temperature with added beliefs is an output of the simulation protocol rather than a tautological re-expression of the model equations.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The model rests on a small set of control parameters and structural assumptions about belief states and network connectivity; no new particles or forces are postulated.

free parameters (2)
  • neutrality parameter
    Tunes the relative conviction of the neutral stance in the three-state belief model.
  • temperature
    Represents external social agitation and tolerance for internal cognitive dissonance.
axioms (2)
  • domain assumption Each belief can take one of two opposite polarized states or a neutral state.
    Core definition of the three-state opinion model invoked throughout the abstract.
  • domain assumption The external social graph is fully connected.
    Stated as the setting for analytical and simulation results.

pith-pipeline@v0.9.0 · 5761 in / 1398 out tokens · 59942 ms · 2026-05-21T02:09:16.274266+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

55 extracted references · 55 canonical work pages

  1. [1]

    Our aim is to understand how the critical and tricritical points vary with system parameters and with the beliefs’ topol- ogy

    We choose two values for the neutrality parameter: α= 0, for which we expect a second-order phase transi- tion, andα= 0.85, which lies in the first-order transition regime for the fully connected internal graph. Our aim is to understand how the critical and tricritical points vary with system parameters and with the beliefs’ topol- ogy. Since, for both th...

    work page

  2. [2]

    Results for α= 0 When α= 0 , all topologies exhibit a second-order phase transition for the number of internal beliefs stud- ied. As observed for ER graphs [32], the MFA under- estimates the critical temperature value, yet it qualita- tively captures the influence of additional beliefs on it for the clique and ring internal topologies, as shown in Fig. 4(...

    work page

  3. [3]

    Results for α= 0.85 When all nodes have the same degreeki = z, which holds true for both ring and clique agents, all beliefs share the same magnetization such thatmµ= m and the fraction of neutral beliefs nµ 0 = n0 for all µ∈ {A,B,C,...}. Consequently, the mean-field free-energy functionL(m,n,β)≡(HMF−β−1SMF)/N simplifies to: L(m,n,β) =−z 2 ( m2 +α2n2) +zα...

    work page 2000

  4. [4]

    ESF Investing in your future

    The values for N0,int are no longer merely thermal fluctuations, but complementMint, meaning that when Mint = 0, it is not due to an equal number of opposite be- liefs, but because there is neutral internal agreement. In this scenario, rings, rather than stars, are the dominant structure. When cliques are alone, they display an almost flat distribution fo...

    work page doi:10.13039/501100011033

  5. [5]

    Axelrod, J

    R. Axelrod, J. Conflict Resolut.41, 203 (1997)

    work page 1997

  6. [6]

    Baumann, P

    F. Baumann, P. Lorenz-Spreen, I. M. Sokolov, and M. Starnini, Phys. Rev. X11, 011012 (2021)

    work page 2021

  7. [7]

    T. M. Pham, J. Korbel, R. Hanel, and S. Thurner, Proc. Natl. Acad. Sci. U.S.A. 119, 10.1073/pnas.2121103119 (2022)

    work page doi:10.1073/pnas.2121103119 2022

  8. [8]

    Korbel, S

    J. Korbel, S. D. Lindner, T. M. Pham, R. Hanel, and S. Thurner, Phys. Rev. Lett.130, 057401 (2023)

    work page 2023

  9. [9]

    R. P. Abelson and M. J. Rosenberg, Behav. Sci. 3, 1 (1958)

    work page 1958

  10. [10]

    Hinkov, D

    J. Dalege and T. van der Does, Sci. Adv.8, 10.1126/sci- adv.abm0137 (2022)

    work page doi:10.1126/sci- 2022

  11. [11]

    Vlasceanu, A

    M. Vlasceanu, A. M. Dyckovsky, and A. Coman, Per- spect. Psychol. Sci.19, 444–453 (2023)

    work page 2023

  12. [12]

    Sayama, J

    H. Sayama, J. Complex Netw.6, 430–447 (2017)

    work page 2017

  13. [13]

    Sporns, G

    O. Sporns, G. Tononi, and R. Kötter, PLoS Comput. Biol. 1, e42 (2005)

    work page 2005

  14. [14]

    A. P. Davison and S. Appukuttan, eLife 11, e84463 (2022)

    work page 2022

  15. [15]

    Zitnik and J

    M. Zitnik and J. Leskovec, Bioinformatics33, i190–i198 (2017)

    work page 2017

  16. [16]

    H. Kim, O. K. Pineda, and C. Gershenson, Complexity 2019, 1–11 (2019)

    work page 2019

  17. [17]

    L. A. Escobar, H. Kim, and C. Gershenson, Complexity 2019, 1–14 (2019)

    work page 2019

  18. [18]

    M. S. Granovetter, Am. J. Sociol.78, 1360 (1973)

    work page 1973

  19. [19]

    N. E. Friedkin, A. V. Proskurnikov, R. Tempo, and S. E. Parsegov, Science 354, 321 (2016)

    work page 2016

  20. [20]

    H. L. J. van der Maas, J. Dalege, and L. Waldorp, J. Complex Netw. 8, 10.1093/comnet/cnaa010 (2020)

    work page doi:10.1093/comnet/cnaa010 2020

  21. [21]

    Ellinas, N

    C. Ellinas, N. Allan, and A. Johansson, PLoS ONE12, e0180193 (2017)

    work page 2017

  22. [22]

    Dalege, M

    J. Dalege, M. Galesic, and H. Olsson, Psychol. Rev.132, 253–290 (2025)

    work page 2025

  23. [23]

    Bizyaeva, A

    A. Bizyaeva, A. Franci, and N. E. Leonard, IEEE Trans. Autom. Control 70, 5082–5097 (2025)

    work page 2025

  24. [24]

    Castellano, S

    C. Castellano, S. Fortunato, and V. Loreto, Rev. Mod. Phys. 81, 591–646 (2009)

    work page 2009

  25. [25]

    Abelson, in Contributions to mathematical psychol- ogy, edited by N

    R. Abelson, in Contributions to mathematical psychol- ogy, edited by N. Fredericksen and H. Gullicksen (Holt, Rinehart & Winston, New York, 1964)

    work page 1964

  26. [26]

    Deffuant, D

    G. Deffuant, D. Neau, F. Amblard, and G. Weisbuch, Adv. in Compl. Syst.3, 87 (2000)

    work page 2000

  27. [27]

    Deffuant, F

    G. Deffuant, F. Amblard, G. Weisbuch, and T. Faure, J. Artif. Soc. Soc. Simul.5 (2002)

    work page 2002

  28. [28]

    R. A. Holley and T. M. Liggett, Ann. Probab.3, 643 (1975)

    work page 1975

  29. [29]

    Ising, Zeitschrift für Physik31, 253 (1925)

    E. Ising, Zeitschrift für Physik31, 253 (1925)

    work page 1925

  30. [30]

    Galam, Physica A238, 66 (1997)

    S. Galam, Physica A238, 66 (1997)

    work page 1997

  31. [31]

    Sznajd-Weron and J

    K. Sznajd-Weron and J. Sznajd, Int. J. Mod. Phys. C11, 1157–1165 (2000)

    work page 2000

  32. [32]

    Blume, V

    M. Blume, V. J. Emery, and R. B. Griffiths, Phys. Rev. A 4, 1071 (1971)

    work page 1971

  33. [33]

    Yang, Physics Procedia3, 1839 (2010), the Inter- national Conference on Complexity and Interdisciplinary Sciences.The3rdChina-EuropeSummerSchoolonCom- plexity Sciences

    Y.-H. Yang, Physics Procedia3, 1839 (2010), the Inter- national Conference on Complexity and Interdisciplinary Sciences.The3rdChina-EuropeSummerSchoolonCom- plexity Sciences

    work page 2010

  34. [34]

    T. C. Schelling, J. Math. Sociol.1, 143 (1971)

    work page 1971

  35. [35]

    Gauvin, J.-P

    L. Gauvin, J.-P. Nadal, and J. Vannimenus, Phys. Rev. E 81, 066120 (2010)

    work page 2010

  36. [36]

    Ferri, A

    I. Ferri, A. Díaz-Guilera, and M. Palassini, Equilibrium and dynamics of a three-state opinion model (2022), arXiv:2210.03054 [cond-mat.stat-mech]

    work page arXiv 2022

  37. [37]

    Ferri, C

    I. Ferri, C. Pérez-Vicente, M. Palassini, and A. Díaz- Guilera, Entropy 24, 10.3390/e24111627 (2022)

    work page doi:10.3390/e24111627 2022

  38. [38]

    Ferri, A

    I. Ferri, A. Gaya-Àvila, and A. Díaz-Guilera, Chaos33, 093121 (2023)

    work page 2023

  39. [39]

    L.Festinger, A Theory of Cognitive Dissonance (Stanford University Press, Stanford, CA, 1957)

    work page 1957

  40. [40]

    Harmon-Jones,Cognitive Dissonance: Reexamining a Pivotal Theory in Psychology , 2nd ed

    E. Harmon-Jones,Cognitive Dissonance: Reexamining a Pivotal Theory in Psychology , 2nd ed. (American Psy- chological Association, 2019)

    work page 2019

  41. [41]

    McPherson, L

    M. McPherson, L. Smith-Lovin, and J. M. Cook, Annu. Rev. Sociol. 27, 415–444 (2001)

    work page 2001

  42. [42]

    Rodriguez, J

    N. Rodriguez, J. Bollen, and Y.-Y. Ahn, PLOS ONE11, e0165910 (2016)

    work page 2016

  43. [43]

    Hoffer,The True Believer: Thoughts on the Nature of Mass Movements (Harper and Row, 1951)

    E. Hoffer,The True Believer: Thoughts on the Nature of Mass Movements (Harper and Row, 1951)

    work page 1951

  44. [44]

    Kovács, G

    Z. Kovács, G. Palla, and A. Zafeiris, Sci. Rep. 16, 10.1038/s41598-025-31643-5 (2025)

    work page doi:10.1038/s41598-025-31643-5 2025

  45. [45]

    J. R. Zaller, The Nature and Origins of Mass Opinion (Cambridge University Press, Cambridge, 1992)

    work page 1992

  46. [46]

    Baldassarri and P

    D. Baldassarri and P. Bearman, Am. Sociol. Rev.72, 784 (2007)

    work page 2007

  47. [47]

    Dalege, D

    J. Dalege, D. Borsboom, F. van Harreveld, and H. L. J. van der Maas, Soc. Psychol. Personal. Sci.8, 528 (2017)

    work page 2017

  48. [48]

    Borsboom, M

    D. Borsboom, M. K. Deserno, M. Rhemtulla, S. Ep- skamp, E. I. Fried, R. J. McNally, D. J. Robinaugh, M. Perugini, J. Dalege, G. Costantini, A.-M. Isvoranu, A. C. Wysocki, C. D. van Borkulo, R. van Bork, and L. J. Waldorp, Nat. Rev. Methods Primers1, 58 (2021)

    work page 2021

  49. [49]

    Gómez-Gardeñes, M

    J. Gómez-Gardeñes, M. de Domenico, G. Gutiérrez, A. Arenas, and S. Gómez, Philos. Trans. R. Soc. A373, 20150117 (2015)

    work page 2015

  50. [50]

    S. Jang, J. S. Lee, S. Hwang, and B. Kahng, Phys. Rev. E 92, 022110 (2015)

    work page 2015

  51. [51]

    Chmiel, J

    A. Chmiel, J. Sienkiewicz, and K. Sznajd-Weron, Phys. Rev. E 96, 062137 (2017)

    work page 2017

  52. [52]

    Krawiecki, Physica A506, 773 (2018)

    A. Krawiecki, Physica A506, 773 (2018)

    work page 2018

  53. [53]

    Zhang, C

    J. Zhang, C. Li, and J. Wang, Biometrics79, 3564–3573 (2023). 15

    work page 2023

  54. [54]

    Arenas, A

    A. Arenas, A. Díaz-Guilera, and C. J. Pérez-Vicente, Phys. Rev. Lett.96, 114102 (2006)

    work page 2006

  55. [55]

    Sales-Pardo, R

    M. Sales-Pardo, R. Guimerà, A. A. Moreira, and L. A. N. Amaral, Proc. Natl. Acad. Sci. U.S.A.104, 15224 (2007)

    work page 2007