{"id":"3263f09b-3111-4331-9021-25dd9409cca0","arxiv_id":"2607.05932","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"Fully heterogeneous binary network games reduce to equivalent three-archetype games (conformist, rebel, stubborn), yielding equilibrium existence conditions, dynamics results, and frequency predictions from network homophily.","lead":"Any binary-choice network game with fully different player payoffs can be rewritten as an equivalent game of only three types—conformists, rebels, and stubborn agents—while keeping the same equilibria and best-response paths. The framework gives existence and convergence conditions, predicts strategy frequencies from network mixing patterns, and is checked on real social networks with heterogeneous altruism.","discovery_kind":"unification","skeptic_critique":{"model":"grok-4.5","headline":"The three-archetype reduction is load-bearing on a restricted payoff class (threshold BR in neighbor counts/weights), not arbitrary heterogeneous binary network payoffs as the abstract wording suggests.","rationale":"The reader’s weakest_assumption is exactly the load-bearing hinge: the reduction and all downstream existence/dynamics/approximation results presuppose a structural form (additive separability + linear or monotone dependence on neighbors) that the abstract does not state. That is not an external-consensus objection; it is an internal scope condition. Without the full-text definition confirming the class, soundness remains unverifiable, so the UNVERDICTED / LOW-confidence posture is still correct. No stronger internal inconsistency is visible from the abstract alone, and the empirical PD check on real networks is consistent with the restricted class. A clean Definition 2.1 of the standard linear network-game form would raise confidence and could move the verdict to CONDITIONAL or ACCEPT; the concrete test above settles it in one step. I therefore leave the reader’s verdict unchanged and mark full agreement on the critical assumption.","tokens_in":2066,"tokens_out":688,"duration_ms":25000,"concrete_test":"Locate the formal definition of the game (likely §2). If payoffs are of the form u_i = a_i·(α_i + ∑_j g_{ij} β_{ij} a_j) + γ_i(a_{N_i}) with β monotone, or BR is required to be threshold in a weighted neighbor sum, the claim is standard and holds. If u_i is an arbitrary function of the neighbor action vector, construct a 2-neighbor player whose BR is XOR of the two neighbors and check whether any assignment of conformist/rebel/stubborn types (with any thresholds) reproduces that BR on all four configurations; if not, the reduction fails and the subsequent theorems shrink to the restricted class.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that any binary-choice network game with fully heterogeneous payoffs transforms into an equivalent conformist/rebel/stubborn game preserving pure NE and best-response trajectories—holds only if each player’s best response is a threshold rule on a one-dimensional statistic of neighbors’ actions (typically a weighted sum or count). Conformist/rebel/stubborn archetypes are essentially monotone threshold (or constant) maps of that statistic. If the maintained class allows arbitrary u_i(a_i, a_{N_i}) : {0,1}×{0,1}^{d_i}→ℝ, then BR can be any Boolean function of the neighbor configuration. Many such functions (e.g., parity of two specific neighbors, or non-monotone pattern matching) cannot be realized by any fixed archetype of the three types, so neither the equilibrium set nor the BR trajectories are preserved. The abstract’s “any such game” / “fully heterogeneous payoff structures” language does not display this restriction; the reader correctly flagged it as unstated. Existence, vanishing, approximation, and limited-information results all inherit the same scope. Without an explicit Definition that pins payoffs to additive/linear or monotone-in-sum form, the reduction is not general.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The manuscript develops a framework for binary-choice network games with fully heterogeneous payoffs. Its central claim is that any such game can be transformed into an equivalent game whose players are only three archetypes—conformist, rebel, and stubborn—while preserving the set of pure-strategy Nash equilibria and best-response trajectories. Building on that reduction, the paper supplies sufficient conditions for pure NE existence and BR-dynamics convergence on arbitrary networks, proves almost-sure vanishing of equilibria on large sparse random graphs, constructs a deterministic approximation that predicts strategy frequencies from homophily/heterophily statistics without computing equilibria, and extends the analysis to limited-information settings (convergence to a unique limited-information equilibrium or unique stationary distribution, with necessary and sufficient existence conditions). Predictions are illustrated on Prisoner's Dilemma games played on real social networks with heterogeneous altruism and peer influence.","tokens_in":2288,"tokens_out":1332,"duration_ms":29726,"significance":"If the three-archetype reduction is correctly scoped and the subsequent theorems hold, the paper would supply a genuine unification: existence, dynamics, vanishing, and outcome prediction for heterogeneous binary network games under a single reduction. The deterministic approximation from network assortativity patterns and the limited-information characterization are potentially useful tools for applied work. The empirical check on real networks with heterogeneous altruism is a welcome external validation. These contributions would matter for the network-games literature, which has largely stayed with homogeneous or few-type preferences.","major_comments":[{"comment":"The load-bearing claim (abstract and opening of the reduction section) that 'any' binary-choice network game with 'fully heterogeneous payoff structures' reduces to a conformist/rebel/stubborn game while preserving pure NE and BR trajectories is not true for arbitrary payoffs u_i(a_i, a_{N_i}):{0,1}×{0,1}^{d_i}→ℝ. Under that unrestricted domain the best response can be an arbitrary Boolean function of the neighbor configuration (parity of two specific neighbors, non-monotone pattern matching, etc.). Conformist/rebel/stubborn archetypes realize only threshold (or constant) maps of a one-dimensional statistic of neighbors' actions. The reduction therefore holds only inside a restricted class—essentially additive/linear or monotone-in-weighted-sum payoffs that induce threshold BR. The manuscript must state an explicit Definition of the maintained payoff class and rewrite every subsequent cl","section":"Abstract; reduction theorem / §2–3"},{"comment":"Existence and BR-convergence theorems inherit the same scope restriction. If the paper's proofs rely on monotonicity or single-crossing of best responses (standard for threshold games), those arguments do not extend to non-monotone Boolean BR. The statements must either (i) list the precise structural assumptions used in the proofs or (ii) supply counter-examples showing where the claims fail outside the threshold class, so that the reader can see the boundary of the results.","section":"Existence / BR-dynamics theorems"},{"comment":"The almost-sure vanishing result on large sparse random networks and the deterministic approximation from homophily/heterophily both treat strategy frequencies as driven by local neighbor counts or weighted sums. That structure is again native to threshold BR. Without an explicit error analysis or a statement that the approximation is for the reduced three-archetype game only, the predictive claims over-reach. A short proposition bounding the approximation error (or a clear caveat that the map is exact only after the reduction) is needed.","section":"Vanishing / deterministic approximation sections"},{"comment":"The limited-information extension asserts convergence to a unique limited-information equilibrium or unique stationary distribution, plus necessary and sufficient existence conditions. These results again presuppose that each agent's information-contingent BR remains a threshold rule of the three-archetype form. If agents receive only coarse signals, non-monotone payoffs can produce multiple stationary distributions even under the paper's information structure. The necessary-and-sufficient conditions should be restated with the maintained payoff class made explicit, and the uniqueness argument should flag where monotonicity is used.","section":"Limited-information theorems"}],"minor_comments":[{"comment":"The abstract's phrasing 'any such game' / 'fully heterogeneous payoff structures' should be aligned with the (to-be-added) formal Definition so that the abstract does not over-claim relative to the body.","section":"Abstract"},{"comment":"Notation for the three archetypes (conformist / rebel / stubborn) should be introduced with explicit best-response maps (e.g., BR = 1 iff weighted neighbor sum ≥ θ, BR = 1 iff weighted sum ≤ θ, BR constant) so that the reduction map is immediately checkable.","section":"Reduction section"},{"comment":"The empirical Prisoner's Dilemma validation on real networks is useful; a short table or figure reporting predicted versus observed strategy frequencies (and the homophily statistics used) would make the external check easier to evaluate.","section":"Empirical section"},{"comment":"Cross-references between the reduction theorem and each later theorem would help the reader track which results are exact consequences of the three-archetype form and which are additional.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The reader's and skeptic's concern about payoff-class scope is the decisive issue; it is fixable by tightening definitions and restating theorems, so major_revision rather than reject is appropriate. Novelty relative to the large literature on threshold/linear network games (Ballester–Calvó-Armengol–Zenou, Jackson–Yariv, etc.) should be checked carefully once the class is made explicit—if the contribution is mainly the packaging into three archetypes plus the vanishing/approximation/limited-info package, that is still publishable but the abstract should not advertise unrestricted heterogeneity. Fit for a serious theory journal is adequate if the revised statements are sharp."},"author_rebuttal":{"model":"grok-4.5","summary":"We thank the referee for a careful and constructive report. The central concern—that the three-archetype reduction and all subsequent theorems are correctly scoped only to games whose best responses are threshold rules of a one-dimensional neighbor statistic, not to unrestricted Boolean best responses—is well taken. We agree that the present wording (“any” binary-choice network game with “fully heterogeneous payoff structures”) overstates the domain. In the revision we will introduce an explicit Definition of the maintained payoff class (payoffs linear or monotone in a weighted sum of neighbors’ actions, inducing threshold BR), rewrite the abstract and every theorem statement to reference that class, add a boundary counter-example, supply an approximation-error bound, and flag where monotonicity is used in the limited-information uniqueness arguments. The technical content of the proofs is unchanged; the claims will simply be correctly scoped. We address each major comment below.","responses":[{"response":"We agree. The reduction we establish applies to binary-action network games in which each player’s payoff difference is a (weakly) monotone function of a weighted sum of neighbors’ actions—equivalently, best responses are threshold rules with respect to that statistic. Within this class (which covers the linear and linear-quadratic specifications standard in the network-games literature, as well as the heterogeneous-altruism Prisoner's Dilemma applications we study), the transformation to conformist, rebel, and stubborn archetypes preserves pure-strategy Nash equilibria and best-response trajectories. Outside this class, best responses can be arbitrary Boolean functions and the reduction fails. In the revision we will: (i) add an explicit Definition of the maintained payoff class at the opening of Section 2; (ii) replace every unqualified “any”/“fully heterogeneous” claim in the abstract, introduction, and theorem statements with a precise reference to that class; (iii) note that the three archetypes exhaust the threshold BR maps obtainable by action-label flips and threshold shifts. These changes correctly scope the contribution without altering the proofs.","revision_made":"yes","referee_comment":"The load-bearing claim (abstract and opening of the reduction section) that 'any' binary-choice network game with 'fully heterogeneous payoff structures' reduces to a conformist/rebel/stubborn game while preserving pure NE and BR trajectories is not true for arbitrary payoffs u_i(a_i,a_{N_i}). Under that unrestricted domain the best response can be an arbitrary Boolean function. Conformist/rebel/stubborn archetypes realize only threshold (or constant) maps of a one-dimensional statistic. The reduction therefore holds only inside a restricted class—essentially additive/linear or monotone-in-weighted-sum payoffs that induce threshold BR. The manuscript must state an explicit Definition of the maintained payoff class and rewrite every subsequent claim."},{"response":"We agree that the existence and BR-convergence results rely on the threshold (monotone single-crossing) structure of best responses and do not extend to arbitrary Boolean BR. Our proofs use the fact that, after the reduction, each player’s best response is a threshold rule in the number (or weighted sum) of neighbors choosing action 1; this yields the potential-function or acyclic-improvement arguments that deliver pure NE existence under the stated network/payoff conditions and global convergence of sequential BR dynamics. We will revise the theorem statements to list these structural assumptions explicitly (the maintained payoff class of the new Definition together with the network conditions already stated). We will also add a brief remark, with a simple counter-example (e.g., a two-neighbor XOR best response on a small cycle), indicating that pure NE need not exist and BR dynamics need not converge once one leaves the threshold class. This makes the boundary of the results transparent.","revision_made":"yes","referee_comment":"Existence and BR-convergence theorems inherit the same scope restriction. If the paper's proofs rely on monotonicity or single-crossing of best responses (standard for threshold games), those arguments do not extend to non-monotone Boolean BR. The statements must either (i) list the precise structural assumptions used in the proofs or (ii) supply counter-examples showing where the claims fail outside the threshold class, so that the reader can see the boundary of the results."},{"response":"We agree that both the vanishing result and the deterministic approximation are native to the reduced three-archetype (threshold) game, in which local frequencies and weighted neighbor counts fully determine best responses. In the revision we will: (i) state at the outset of those sections that the results apply to the reduced game (equivalently, to the maintained payoff class); (ii) add a short proposition that bounds the approximation error between the deterministic frequency map and the expected frequencies under the stochastic BR process on finite networks, under standard concentration assumptions on the degree and type distributions; (iii) clarify that the map is exact in the large-population limit after the reduction, and that the homophily/heterophily statistics enter precisely because they determine the local type-weighted neighbor counts that drive threshold BR. These additions will prevent over-reach while preserving the applied usefulness of the approximation for the class of games we study.","revision_made":"yes","referee_comment":"The almost-sure vanishing result on large sparse random networks and the deterministic approximation from homophily/heterophily both treat strategy frequencies as driven by local neighbor counts or weighted sums. That structure is again native to threshold BR. Without an explicit error analysis or a statement that the approximation is for the reduced three-archetype game only, the predictive claims over-reach. A short proposition bounding the approximation error (or a clear caveat that the map is exact only after the reduction) is needed."},{"response":"We agree. The limited-information results rely on the same threshold structure: under the maintained payoff class, each agent’s information-contingent best response remains a threshold rule of the three-archetype form, which is what delivers uniqueness of the limited-information equilibrium (or of the stationary distribution of the induced Markov chain) and the necessary-and-sufficient existence conditions we derive. With non-monotone Boolean best responses, multiple stationary distributions can arise even under our information structure. In the revision we will restate the theorems with the maintained payoff class made explicit, and we will flag in the uniqueness proofs the precise steps that use monotonicity/single-crossing of the information-contingent best responses. This will correctly delimit the results.","revision_made":"yes","referee_comment":"The limited-information extension asserts convergence to a unique limited-information equilibrium or unique stationary distribution, plus necessary and sufficient existence conditions. These results again presuppose that each agent's information-contingent BR remains a threshold rule of the three-archetype form. If agents receive only coarse signals, non-monotone payoffs can produce multiple stationary distributions even under the paper's information structure. The necessary-and-sufficient conditions should be restated with the maintained payoff class made explicit, and the uniqueness argument should flag where monotonicity is used."}],"tokens_in":2007,"tokens_out":1524,"duration_ms":63300,"standing_objections":[]},"desk_editor":{"model":"grok-4.5","letter":"Punchline: they give a clean organizing map that turns fully heterogeneous binary network games into equivalent games with only three player types—conformist, rebel, stubborn—while preserving pure NE and best-response trajectories, then stack existence, BR convergence, a vanishing result on large sparse random graphs, a homophily-based frequency predictor, and a limited-information extension on top. That is a real contribution inside network game theory if the proofs hold.\n\nWhat is actually new is the claim that the reduction is general within their class and preserves paths, not just equilibria, plus the package of existence/vanishing/approximation results under full preference heterogeneity. Most of the literature still works with homogeneous players or a handful of types. The deterministic approximation that reads equilibrium frequencies off network homophily/heterophily without solving for equilibria is practically useful. The real-network PD check with heterogeneous altruism and peer influence is a sensible validation step for a theory paper, not window dressing.\n\nSoft spot, in proportion: the reduction is load-bearing on best responses being threshold rules on a one-dimensional neighbor statistic (weighted sum or count). That is the standard network-game payoff form, so the paper is almost certainly fine inside the literature it sits in. But the abstract’s “any such game” / “fully heterogeneous payoff structures” wording does not display that restriction. If payoffs allow arbitrary Boolean functions of the neighbor configuration—parity of two neighbors, non-monotone pattern matching—the three archetypes cannot realize the BR map, and neither equilibria nor trajectories are preserved. Existence, vanishing, approximation, and limited-info results inherit the same scope. Minor if Definition 1 pins payoffs to additive/monotone-in-sum form; more serious if the body keeps the abstract’s unrestricted language. I have not seen a load-bearing circularity or fitting-as-prediction problem on the visible program.\n\nWho it is for: theorists and applied modelers who need existence and frequency tools under preference heterogeneity on networks. It deserves a serious referee. Send it out; expect the referee to demand a crisp payoff-class definition and a clear statement of what “any” means. I would bring it to a reading group if someone in the room works on network games.","headline":"Useful three-archetype reduction and dynamics package for heterogeneous binary network games, but the abstract oversells generality relative to the usual threshold-payoff class.","tokens_in":2924,"tokens_out":551,"would_cite":false,"duration_ms":24011,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["91A43","91D30"],"pacs":[],"model":"grok-4.5","headline":"Any binary-choice network game with fully heterogeneous payoffs reduces to three player archetypes while preserving equilibria and best-response paths.","keywords":["network games","heterogeneous players","binary choice","Nash equilibrium","best-response dynamics","player archetypes","homophily","limited information"],"falsifier":"Exhibit a binary network game whose payoffs are not separable into own-action plus monotone neighbor effects (for example a payoff that rewards a specific three-player motif) and check whether the claimed three-archetype transformation still produces identical pure Nash equilibria and best-response trajectories; failure of the match falsifies the claimed generality of the reduction.","tokens_in":2910,"feed_emoji":"🕸️","tokens_out":951,"duration_ms":28936,"temperature":0.7,"pith_summary":"This paper shows that binary-choice games played on networks by players with completely different payoff structures can always be rewritten as an equivalent game that uses only three types of player: conformists, who prefer to match their neighbors; rebels, who prefer to differ from them; and stubborn players, whose preferred action does not depend on the network. The rewriting leaves the set of pure-strategy Nash equilibria and the entire trajectory of best-response dynamics unchanged. From the reduced game the authors derive sufficient conditions for pure equilibria to exist and for best-response dynamics to converge on any network, while proving that pure equilibria disappear with probability one in large sparse random networks. They also give a deterministic approximation that forecasts long-run strategy frequencies from the network’s homophily and heterophily statistics alone, without solving for equilibria. When players have only limited information the same reduction yields a unique limited-information equilibrium or a unique stationary distribution, and the predictions are checked on Prisoner's Dilemma games with heterogeneous altruism and peer influence played on real social networks.","feed_headline":"Any network game reduces to three player archetypes","feed_subtitle":"The map preserves equilibria and dynamics, yielding existence results and homophily-based forecasts","key_machinery":"The three-archetype payoff transformation: a map that rewrites each player’s payoff so that the player becomes either a pure conformist, a pure rebel, or a stubborn type whose preferred action is independent of neighbors, while leaving the best-reply correspondence and therefore the pure Nash set and best-response paths invariant.","core_discovery":"Any binary-choice network game with arbitrary heterogeneous payoffs can be transformed into an equivalent game whose players belong to only three archetypes—conformist, rebel, and stubborn—so that the pure-strategy Nash equilibria and the best-response trajectories of the original game are exactly preserved.","pith_inferences":["The three-archetype classification supplies a low-dimensional typology that empirical network studies could use in place of continuous preference estimation.","Almost-sure non-existence of pure equilibria in sparse random graphs implies that observed coordination on large online platforms must rely on mixed strategies, denser local structure, or learning rules other than pure best response.","The homophily-based deterministic approximation can be tested as a reduced-form predictor of adoption or opinion cascades without full network simulation.","If the reduction truly requires monotone neighborhood effects, games with higher-order interactions would need either a larger archetype set or an entirely different reduction."],"forward_implications":["Pure-strategy Nash equilibria exist on arbitrary networks whenever the sufficient conditions obtained from the reduced three-type game are met.","Best-response dynamics converge on any network under those same conditions.","In large sparse random networks pure equilibria vanish with probability approaching one.","Long-run strategy frequencies and evolutionary trends can be read off from the network’s homophily and heterophily patterns without computing equilibria.","Under limited information the dynamics converge either to a unique limited-information equilibrium or to a unique stationary distribution."],"fun_headline_variants":["Any network game maps to three archetypes","Heterogeneous payoffs reduce to conformist rebel stubborn","Binary network games collapse into three player types","Full payoff diversity equals three archetypes preserving equilibria","Network games transform into three simple player archetypes"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The reduction and all subsequent existence and dynamics results rest on the standard structural form of network-game payoffs—each player’s payoff is additively separable in own action and a monotone function of neighbors’ actions; if payoffs involve non-monotone or higher-order motif interactions the three-type map need not preserve the game.","fun_headline_variants_meta":{"raw":{"variants":["Any network game maps to three archetypes","Heterogeneous payoffs reduce to conformist rebel stubborn","Binary network games collapse into three player types","Full payoff diversity equals three archetypes preserving equilibria","Network games transform into three simple player archetypes"]},"model":"grok-4.5","cost_usd":0.008076,"raw_usage":{"total_tokens":1856,"prompt_tokens":717,"num_sources_used":0,"completion_tokens":72,"cost_in_usd_ticks":80760000,"prompt_tokens_details":{"text_tokens":717,"audio_tokens":0,"image_tokens":0,"cached_tokens":128},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":1067,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":717,"tokens_out":72,"duration_ms":12164,"temperature":1.0,"reasoning_tokens":1067,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-08T19:26:00.369166+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Exhibit a binary network game whose payoffs are not separable into own-action plus monotone neighbor effects (for example a payoff that rewards a specific three-player motif) and check whether the claimed three-archetype transformation still produces identical pure Nash equilibria and best-response trajectories; failure of the match falsifies the claimed generality of the reduction.","supporting_citations":[],"review_version":1}