{"id":"f955ff56-d73c-4ece-ad4f-21cce29d5c2f","arxiv_id":"1908.00660","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A homogeneous pair approximation for the q-voter model with independence on two-layer multiplex networks accurately predicts transition points and hysteresis widths for dense, homogeneous layers, but fails for sparse or strongly heterogeneous layers.","lead":"This paper extends a standard approximation method for opinion-spreading models to networks made of multiple connected layers, and checks it against computer simulations. It shows the approximation works well when each layer has many connections per person, and fails when connections are sparse or highly uneven.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the central quantitative-agreement claim is parameter-free, internally consistent, and explicitly scoped; the main issues are presentation-level (missing error bars/code and a Table I caption typo).","rationale":"The reader's weakest assumption points to the binomial independence and mean-degree-only reduction in the homogeneous PA. I agree that this is the approximation most likely to limit the theory, but it is not a hidden flaw: the paper explicitly identifies this as the source of failure for strongly heterogeneous layers and for small mean degree, and the central claim is carefully restricted to the regime where the approximation is expected to work. The derivation of the PA is parameter-free, and the analytic critical-point formulas in the Appendix are consistent with the MC values reported for homogeneous layers when the correct q is used. The only substantive issue I found is a data-labeling inconsistency: Table I and Fig. 1 are captioned q=4, yet the numerical entries match the PA for q=3, which means those particular exponent and pc values should not be cited as evidence about the q=4 case without correction. This is a presentation/reproducibility problem rather than a threat to the main quantitative-agreement claim, so the reader's conditional verdict remains appropriate; no change in verdict is needed, though the authors should fix the caption and ideally release data/code with error bars.","tokens_in":26490,"tokens_out":15591,"duration_ms":157551,"concrete_test":"Recompute the PA critical values from Eqs. (47)-(48) for the network parameters listed in Table I (RRG <k>=10; SF lambda=3.0, kmin=10; SF lambda=2.5, kmin=10) for q=3 and q=4, and compare with the tabulated pc values; the entries should reproduce q=3, confirming the caption typo. If the authors instead intend q=4, the PA-vs-MC comparison in Table I would contradict the central claim and must be re-examined.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that the homogeneous PA be accurate for dense, homogeneous or weakly heterogeneous duplex layers. The weakest step is Eq. (19): independent binomial active-bond counts and degree-independent theta_j, which reduces all degree-distribution information to <k>. This is a real approximation, but the paper does not overclaim: it explicitly restricts the quantitative-agreement claim to <k> >> q and homogeneous or weakly heterogeneous layers, and it explicitly reports qualitative or wrong predictions for strong SF heterogeneity and for <k> comparable to q. The PA is parameter-free, and its critical formula Eq. (47) is analytically derived; for q=3 on an RRG with <k>=10 it gives p*=0.5152, reproducibly matching Table I. I therefore find no load-bearing correctness objection. One presentational defect: Table I and Fig. 1 are labeled q=4, but all listed values (0.5152, 0.535, 0.541) match the PA and MFA expectations for q=3, not q=4; the label should be corrected before the data are used as evidence about q=4.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper extends the homogeneous pair approximation (PA) to two-state spin models on multiplex networks with fully overlapping layers, and applies it to the q-voter model with independence on duplex networks under the LOCAL&AND and GLOBAL&AND update rules. For two identically distributed layers the PA reduces to a two-variable system (Eqs. (25) and (30)) depending only on the mean degree <k>, and the authors derive closed-form expressions for the instability point of the paramagnetic phase (Eqs. (47)-(50)). The theoretical predictions are compared with Monte Carlo simulations on random regular, Erdős-Rényi, and scale-free layers. The central claim, stated in the abstract, is that for mean degree substantially larger than q the homogeneous PA gives quantitative agreement for homogeneous and weakly heterogeneous layers, qualitative agreement for strongly heterogeneous scale-free layers, and can be qualitatively wrong when <k> is comparable to q.","tokens_in":26724,"tokens_out":4915,"duration_ms":46841,"significance":"If the results hold, this is a useful methodological contribution: it provides a parameter-free analytic treatment of a nontrivial multiplex opinion-dynamics model, with transparent derivations and explicit closed-form critical-point formulas that reproduce the mean-field limit as <k>→∞. The paper is honest about the scope of the approximation, explicitly reporting failure regimes for small <k> and for strongly heterogeneous layers, and it names the uncontrolled assumptions (independent binomial active-bond counts, degree-independent θ_j). The comparison to Monte Carlo data for <k>≫q on RRGs, ERGs, and weakly heterogeneous SF networks is convincing and would be a solid basis for the central claim once the data-labeling issue below is resolved.","major_comments":[{"comment":"The data presented as q=4 are quantitatively inconsistent with q=4 and instead match the q=3 predictions of the paper's own Eqs. (47)-(48). For an RRG with <k>=10, Eq. (48) gives θ*=(10-1)/(20-1)=0.4737, and Eq. (47) for q=3 gives p*=0.5152, exactly the value in Table I; for q=4 the same formula gives p*≈0.413, not 0.5152. Likewise, the values 0.535 and 0.541 for SF λ=3.0 and λ=2.5 with kmin=10 are precisely the q=3 results for <k>=20 and <k>=30. Thus Table I and Fig. 1 do not support the stated q=4 finite-size scaling analysis: either the simulations were actually run with q=3 and the labels are wrong, or the quoted critical values are not those of the model described. This must be corrected and the affected statements in Sec. IV.B about the q=4 transition and its exponents re-examined.","section":"Table I and Fig. 1, Sec. IV.B"}],"minor_comments":[{"comment":"The last panel of Fig. 4 is labeled (e) twice; the panel showing q=6, <k>=7 should be labeled (f), and the in-text references to subpanels (Fig. 3(a,b), etc.) in Sec. IV.C should be checked against the correct figure number.","section":"Fig. 4 caption"},{"comment":"The cross-reference in Sec. III.B to 'Sec. III.B' for the application to the q-voter model should read Sec. III.C, and the opening of Sec. III.C referring to 'Sec. III.A' should read Sec. III.B.","section":"Sec. III.B and Sec. III.C"},{"comment":"The caption spells 'GLOBA&AND'; this should be 'GLOBAL&AND'.","section":"Fig. 3 caption"},{"comment":"The Monte Carlo data are presented without error bars or sample counts; adding these (at least for the phase-diagram points and critical values) would strengthen the quantitative comparison.","section":"Figs. 2-5 and Table I"}],"recommendation":"major_revision","confidential_remarks":"The mislabeling of Table I and Fig. 1 as q=4 when the numbers are q=3 is more than a typo: it affects the evidence for the q=4 transition and must be fixed before publication. I otherwise found the derivations sound and the scoping of the central claim appropriate."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new thing here is the homogeneous pair approximation for a general two-state spin model on multiplex networks, worked out for the q-voter with independence under LOCAL&AND and GLOBAL&AND update rules. The closed critical-point formulas in the Appendix are the strongest part: no fitting, analytically derived, and they reproduce the simulation values for dense homogeneous layers. The comparison with Monte Carlo on RRGs, ERGs and weakly heterogeneous SF networks is convincing. The GLOBAL&AND case being equivalent to the 2q-voter model on an aggregate monoplex network is verified by simulation and explicitly stated. That is a real result worth having.\n\nThe soft spots are exactly where the method is approximate. Eq. (19) assumes independent binomial active-bond counts and degree-independent thetas, so all degree-distribution detail collapses to <k>. The paper is honest about the cost: predictions for strongly heterogeneous SF layers are only qualitative, and for <k> near q they can be qualitatively wrong. That limits the central claim to dense homogeneous layers, and the paper says so in the abstract. No hidden parameter fitting, no circularity. The citation pattern is legitimate reuse of known monoplex PA equations, not a problem.\n\nOne presentational defect is real and should be fixed: Table I and Fig. 1 are labeled q=4, but the pc values listed (0.5152 for RRG k=10, 0.535/0.541 for SF) match the q=3 formulas, not q=4. Using Eq. (47)-(48) with q=3 gives 0.515 for RRG and about 0.54 for those SF parameters; q=4 would give roughly 0.41 and 0.45. So the critical exponents in Table I are presumably for q=3, and the text/caption need correction before the numbers are used as evidence about q=4. Also, MC data have no error bars and no code or data release; for a comparison paper that is a minor reproducibility gap, not a correctness flaw.\n\nOverall: competent, honest extension of a known approximation to multiplex networks, with the failure regimes mapped. It belongs in the opinion-dynamics literature and deserves a serious referee. I'd accept it for review and only ask for the label fix, optional error bars, and a data/code statement.","headline":"A parameter-free homogeneous PA for q-voter dynamics on duplex networks, with closed critical-point formulas and honest failure scoping; worth refereeing once the q=3/q=4 label error in Table I and Fig. 1 is fixed.","tokens_in":27200,"tokens_out":2605,"would_cite":true,"duration_ms":26603,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A homogeneous pair approximation derived for two-state spin models on multiplex networks is shown to quantitatively predict the ferromagnetic transition of the q-voter model with independence when the layer mean degree is much larger than…","keywords":["q-voter model","pair approximation","multiplex networks","opinion dynamics","ferromagnetic transition","complex networks","independence noise","duplex networks"],"falsifier":"Simulate the LOCAL&AND q-voter model on two-layer random regular graphs and on homogeneous random graphs with the same mean degree, both with $\\langle k\\rangle\\gg q$, and compare the measured critical p: if the two agree within error, the mean-degree-only closure is supported; if they differ, the claim that only $\\langle k\\rangle$ matters is falsified. Alternatively, measure the empirical distribution of active-bond counts around high-degree versus low-degree nodes in a strongly scale-free layer; a visible deviation from the single-$\\theta$ binomial distribution would isolate the mechanism behind the paper's only-qualitative agreement there.","tokens_in":26327,"feed_emoji":"🧲","tokens_out":9115,"duration_ms":82763,"temperature":0.7,"pith_summary":"This paper establishes a homogeneous pair approximation for any two-state spin model with up-down symmetry on a multiplex network, meaning two or more layers of edges over the same nodes, and applies it to the q-voter model with independence, where agents either follow a unanimous q-neighbor lobby or act independently with probability p. The central aim is to show that this approximation, despite using only the mean degree $\\langle k\\rangle$ of each layer, captures the ferromagnetic transition between disordered and unanimous opinion states. For two layers with identical degree distributions and either the LOCAL&AND or GLOBAL&AND update rule, the theory predicts the order of the transition, the critical values of p, and the width of hysteresis, provided $\\langle k\\rangle$ is substantially larger than q. Monte Carlo simulations confirm quantitative agreement on homogeneous random graphs, random regular graphs, and weakly scale-free layers; agreement is only qualitative on strongly scale-free layers and fails for small $\\langle k\\rangle$ comparable to q.","feed_headline":"Mean degree alone predicts q-voter transitions on two-layer networks","feed_subtitle":"For dense, homogeneous layers one formula matches simulations; for hub-dominated or sparse layers it breaks down.","key_machinery":"The load-bearing object is the homogeneous pair approximation: instead of tracking every degree class, it tracks only the global concentration $c$ of up spins and the concentrations $b^{(A)}$, $b^{(B)}$ of active bonds (edges joining opposite spins) in each layer, and closes the dynamics by assuming that the numbers $i^{(A)}$, $i^{(B)}$ of active bonds around a node are independent binomial variables with degree-independent probabilities $\\theta_j = b/(2c)$ or the analogous expression for the other layer. This closure, together with the adiabatic assumption $c_{k^{(A)},k^{(B)},\\uparrow}\\approx c$, yields the closed two-equation systems Eq. (25) for LOCAL&AND and Eq. (30) for GLOBAL&AND, in which the degree distribution enters only through $\\langle k\\rangle$. The machinery carries the argument because it converts the many-node stochastic process into a low-dimensional dynamical system whose fixed points and stability can be computed analytically, including the critical $p^\\star$ formulas obtained from linear stability analysis of the paramagnetic fixed point.","core_discovery":"The paper's central claim is that the homogeneous pair approximation, developed here for a general two-state spin model with up-down symmetry on a multiplex network, captures the ferromagnetic transition of the q-voter model with independence on two layers with identical degree distributions and full node overlap. The approximation reduces the stochastic many-node dynamics to a two-variable system for the concentration $c$ of up spins and the active-bond concentration $b$, with all degree-distribution information entering only through the mean degree $\\langle k\\rangle$. For the LOCAL&AND update rule, where the spin-flip rate factorizes across layers, the fixed-point analysis predicts continuous transitions for $q=2,3$, a tricritical point at $q=4$, and discontinuous transitions with hysteresis for $q\\geq 5$; for GLOBAL&AND, the system is equivalent to the $2q$-voter model with independence on an aggregate monoplex network of mean degree $2\\langle k\\rangle$, giving continuous transitions for $q=2$ and discontinuous ones for $q\\geq 3$. Linear stability analysis of the paramagnetic fixed point yields the closed-form critical value $p^\\star = 2(2q-1)(\\theta^\\star)^q/[1+2(2q-1)(\\theta^\\star)^q]$ for LOCAL&AND, with $(\\theta^\\star)^q$ replaced by $(\\theta^\\star)^{2q}$ for GLOBAL&AND, where $\\theta^\\star = (\\langle k\\rangle-1)/(2\\langle k\\rangle-1)$, recovering the mean-field result as $\\langle k\\rangle\\to\\infty$. The paper asserts, and supports with Monte Carlo simulations, that these predictions are quantitatively correct when $\\langle k\\rangle\\gg q$ on homogeneous random, random regular, and weakly scale-free layers, only qualitatively correct on strongly scale-free layers, and in general qualitatively wrong when $\\langle k\\rangle$ is small and comparable to $q$.","pith_inferences":["Going beyond the paper: because the homogeneous PA compresses all topology into $\\langle k\\rangle$, it predicts that two-layer networks with identical mean degree but very different degree distributions should show identical critical p and hysteresis width; measuring this directly on, say, bimodal versus regular random graphs would test the limits of the closure.","Going beyond the paper: the same binomial closure suggests that the quantitative failure on strongly scale-free layers should be attributable to degree-dependent active-bond counts, so a degree-resolved heterogeneous PA should restore quantitative accuracy; the paper itself flags this as the natural next step.","Going beyond the paper: the model with small mean degree near q may be a useful testbed for corrections to PA, because the independence assumption is most strained when lobbies are a large fraction of a node's neighborhood and the same neighbor can appear in both layers' lobbies.","Going beyond the paper: the predicted tricritical point at $q=4$ for LOCAL&AND could be located precisely by simulating $q=4$ on random regular graphs across a range of $\\langle k\\rangle$ and checking whether the magnetization jump vanishes continuously at the PA's stability boundary."],"forward_implications":["If the approximation is right, the GLOBAL&AND q-voter model on a two-layer network is dynamically equivalent to the $2q$-voter model with independence on a single aggregate network of mean degree $2\\langle k\\rangle$, a fact confirmed by simulation for all topologies tested.","The closed-form critical values $p^\\star$ for both update rules give the location of the continuous transition, or the lower spinodal of the discontinuous one, directly from $q$ and $\\langle k\\rangle$, with the mean-field complete-graph values recovered as $\\langle k\\rangle\\to\\infty$.","The first-order versus second-order boundary is predicted to lie at $q=4$ for LOCAL&AND, a tricritical point, and between $q=2$ and $q=3$ for GLOBAL&AND, independent of layer topology in the regime $\\langle k\\rangle\\gg q$.","Quantitative agreement is expected only when $\\langle k\\rangle\\gg q$ and the layer degree distributions have finite second moment; strongly scale-free layers with $2<\\lambda<3$ require a heterogeneous pair approximation.","The same pair-approximation framework extends to other binary-state models with up-down symmetry on multiplex networks, such as the q-neighbor Ising model, and to networks with more than two layers."],"supporting_citations":[{"why":"defines the q-voter model with independence on a duplex clique and supplies the MFA spin-flip rates for LOCAL&AND and GLOBAL&AND that the PA must match in the complete-graph limit.","marker":"[16]"},{"why":"derives the pair approximation for the q-voter model with independence on monoplex complex networks, whose binomial closure and summation techniques are extended here to multiplex networks.","marker":"[17]"},{"why":"supplies the stochastic pair approximation treatment of the noisy voter model, including the weighted concentration and active-bond formalism used in the general derivation.","marker":"[20]"},{"why":"provides the analytical and numerical treatment of the nonlinear noisy voter model on complex networks and the critical-exponent relations used to interpret finite-size scaling results.","marker":"[21]"},{"why":"gives the general binary-state pair approximation result that at the critical point $\\theta^\\star=(k-2)/(2(k-1))$ on random regular graphs, which Eqs. (48) and (50) generalize to the duplex case.","marker":"[28]"},{"why":"presents the heterogeneous pair approximation for voter models that the authors identify as the needed upgrade for strongly heterogeneous layers.","marker":"[11]"},{"why":"shows how a heterogeneous pair approximation can be formulated for a binary-state model on multiplex networks, the example the authors cite for extending their homogeneous PA.","marker":"[45]"},{"why":"derives the MFA critical point for the q-voter model with stochastic driving on complete graphs, used as the $\\langle k\\rangle\\to\\infty$ benchmark for the GLOBAL&AND rule.","marker":"[15]"}],"fun_headline_variants":["Q-voter order on two layers is set by mean degree alone","Q-voter pair approximation works only when layers have high mean degree","Homogeneous pair approximation fails for sparse and hub-dominated duplex layers","Two-layer q-voter: theory matches simulation only for dense, homogeneous layers","Q-voter on multiplex networks: mean degree is enough, unless layers are sparse"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The calculation assumes that, around any chosen agent, the number of disagreeing neighbors in each layer is drawn from an independent binomial distribution with a single layer-wide average rate, so every degree-distribution detail beyond the mean degree drops out of the predictions.","fun_headline_variants_meta":{"raw":{"variants":["Q-voter order on two layers is set by mean degree alone","Q-voter pair approximation works only when layers have high mean degree","Homogeneous pair approximation fails for sparse and hub-dominated duplex layers","Two-layer q-voter: theory matches simulation only for dense, homogeneous layers","Q-voter on multiplex networks: mean degree is enough, unless layers are sparse"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000963,"raw_usage":{"total_tokens":4257,"prompt_tokens":1260,"completion_tokens":2997,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":876,"completion_tokens_details":{"reasoning_tokens":2902}},"tokens_in":876,"tokens_out":2997,"duration_ms":21165,"temperature":1.0,"reasoning_tokens":2902,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:39:46.552639+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate the LOCAL&AND q-voter model on two-layer random regular graphs and on homogeneous random graphs with the same mean degree, both with $\\langle k\\rangle\\gg q$, and compare the measured critical p: if the two agree within error, the mean-degree-only closure is supported; if they differ, the claim that only $\\langle k\\rangle$ matters is falsified. Alternatively, measure the empirical distribution of active-bond counts around high-degree versus low-degree nodes in a strongly scale-free layer; a visible deviation from the single-$\\theta$ binomial distribution would isolate the mechanism behind the paper's only-qualitative agreement there.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"derives the pair approximation for the q-voter model with independence on monoplex complex networks, whose binomial closure and summation techniques are extended here to multiplex networks."},{"cited_title":"J¸ edrzejewski, Pair approximation for the q-voter model with independence on complex networks, Phys","cited_arxiv_id":null,"evidence_quote":"supplies the stochastic pair approximation treatment of the noisy voter model, including the weighted concentration and active-bond formalism used in the general derivation."},{"cited_title":"Abramiuk, J","cited_arxiv_id":null,"evidence_quote":"provides the analytical and numerical treatment of the nonlinear noisy voter model on complex networks and the critical-exponent relations used to interpret finite-size scaling results."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"gives the general binary-state pair approximation result that at the critical point $\\theta^\\star=(k-2)/(2(k-1))$ on random regular graphs, which Eqs. (48) and (50) generalize to the duplex case."},{"cited_title":"Krawiecki, Ferromagnetic transition in a simple variant of the Ising model on multiplex networks, Physica A 492, 534 (2018)","cited_arxiv_id":null,"evidence_quote":"shows how a heterogeneous pair approximation can be formulated for a binary-state model on multiplex networks, the example the authors cite for extending their homogeneous PA."},{"cited_title":"Castellano, M","cited_arxiv_id":null,"evidence_quote":"derives the MFA critical point for the q-voter model with stochastic driving on complete graphs, used as the $\\langle k\\rangle\\to\\infty$ benchmark for the GLOBAL&AND rule."}],"review_version":1}