{"id":"278966cd-a2e7-4112-999b-f89c6c2bd0d7","arxiv_id":"1908.05923","paper_version":4,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Correlations between cooperative strategy and node degree enhance cooperation in scale-free networks, and in degree-assorted Poisson networks can make Poisson networks outperform scale-free ones by blocking defector spread at cluster bridges.","lead":"Using computer simulations, the authors show that when cooperators occupy well-connected nodes, cooperation improves in scale-free networks but not in Poisson networks. Adding degree assortativity lets Poisson networks support high cooperation, because bridge areas between degree-based clusters block the spread of defection, unless 'Trojan horse' defectors sit inside cooperator clusters.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central persistence claim is measured at tmax=10^4 with no stationarity check, and the paper's own stepwise-invasion evidence (Figs. 4–5) suggests the bridge barriers may only delay, not prevent, defection.","rationale":"The reader's weakest assumption (the proportional-imitation update rule) is a legitimate robustness concern, but it is not the most load-bearing issue. The proposed barrier mechanism relies primarily on the payoff comparison at cluster interfaces, which would persist under many asynchronous update schemes. The stationarity issue is singled out because the paper's own Figs. 4 and 5 provide direct evidence of non-equilibrium, stepwise invasion dynamics, and the central claim uses words like 'maintain' and 'for all b > 1' that imply persistence. Without a demonstration that the observed cooperation plateau is not a transient of the arbitrarily chosen 10^4-generation window, the headline result is not yet established. The proposed longer-horizon test is cheap and decisive. Since this is a missing verification rather than a demonstrated contradiction, the overall verdict remains CONDITIONAL; because the reader already assigned CONDITIONAL, no change to the verdict category is needed.","tokens_in":12278,"tokens_out":8063,"duration_ms":84809,"concrete_test":"Re-run the degree-assorted Poisson, perfect-correlation, Prisoner's Dilemma cases at b=2 and b=1.2 (and optionally b=1.5) with tmax=10^6, averaging over the last 10^5 steps on the same 50 network realizations. Also record the times of individual cluster invasions. If the tmax=10^6 mean remains at or above the tmax=10^4 value and the invasion staircase stops, the persistence claim is supported. If the mean falls materially (for example below the reported ~20% at b=2), the central claim is a finite-horizon artifact and the abstract's wording should be revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline result is that degree-assorted Poisson networks maintain high cooperation under perfect strategy-degree correlation because bridge areas block defection. The evidence is the mean cooperator fraction over the last 100 of 10^4 timesteps (§II.D and §V.A). The paper never demonstrates that this quantity is stationary. Its own diagnostics point the other way: Fig. 5 shows average defector degree advancing in stair-step jumps ('very rapid steps... separated by long periods of no change'), and Fig. 4 shows final fractions clustered in discrete peaks corresponding to replicates stopped at different bridges. These are signatures of an ongoing, slow invasion process, not a settled equilibrium. The text even says a defector cluster 'might eventually overcome' a bridge. If tmax were extended, more clusters could fall and the reported 'above 20%' plateau could erode, undermining the central 'maintain high levels' claim and the comparison with scale-free networks. The mechanism (degree clustering plus payoff condition (6)) may be real, but the persistence claim is currently horizon-dependent. This is more load-bearing than dependence on update rule (3), since the stationarity question determines whether the reported effect exists beyond the chosen simulation window.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper uses agent-based simulation to study how the initial correlation between an individual's propensity to cooperate and its network degree affects the evolution of cooperation in the Prisoner's Dilemma and Snowdrift games. The authors consider standard Erdős-Rényi (Poisson) and Barabási-Albert (scale-free) networks, as well as degree-assorted versions generated with the Xulvi-Brunet-Sokolov algorithm, and three levels of strategy-degree correlation: none, intermediate (stochastic by degree), and perfect (deterministic assignment of cooperators to the highest-degree nodes). The main finding is that in degree-assorted Poisson networks with perfect strategy-degree correlation, cooperation persists at levels above about 20% for all b>1 in the Prisoner's Dilemma, and can even exceed the cooperation levels of scale-free networks, because bridge areas between clusters of similar degree act as barriers to the spread of defection. This effect is absent under intermediate correlation, which the authors attribute to 'Trojan horses,' i.e., defectors placed within cooperator clusters.","tokens_in":12481,"tokens_out":8134,"duration_ms":76305,"significance":"If the main result is robust, it challenges the common view that scale-free degree heterogeneity is always the most effective network property for promoting cooperation and highlights a novel mechanism—strategy-position correlations interacting with degree assortativity—that could be relevant for interpreting experiments and real-world social networks. The paper is strong in its clarity: the model is standard and precisely specified (Sections II and V), the assignment procedures are explicit, and the simulation protocol is conventional. The authors also provide a useful analytical condition (Eq. 6) for when a cooperator outperforms a defector. However, the central claim about persistence of cooperation rests on a finite simulation horizon without a stationarity check, the main quantitative results lack error bars or statistical tests, and the proposed bridge-barrier mechanism is supported only by qualitative evidence. These are significant limitations that should be addressed before the results can be accepted as stated.","major_comments":[{"comment":"The central persistence claim, that degree-assorted Poisson networks with perfect strategy-degree correlation maintain cooperation above about 20% for all b>1, is measured at tmax=10^4 by averaging the cooperator fraction over the last 100 timesteps, but no stationarity check is reported. The paper's own diagnostics indicate that the system is still evolving: Fig. 4 shows final cooperator fractions clustered in distinct peaks corresponding to defector invasions stopped at different bridges, and Fig. 5 shows the average defector degree increasing in rapid steps separated by long periods of no change. The text in Section IV states that a defector cluster 'might eventually overcome' a bridge. This suggests that the barriers may delay, rather than prevent, the spread of defection, and that the reported plateau above 20% could erode if the simulation were run longer. Please provide a stationarity test (e.g., longer runs, or a quantitative analysis of waiting times at bridges) or demonstrate that the remaining bridges are stable indefinitely.","section":"Sections II.D and V.A; Figs. 2f, 4, 5"},{"comment":"The main quantitative claims—that strategy-degree correlation has no effect in standard Poisson networks, that perfect correlation gives a large enhancement in degree-assorted Poisson networks, and that cooperation in assorted Poisson networks can exceed that in scale-free networks—are based on mean final cooperator fractions over 50 replicates plotted without error bars, confidence intervals, or significance tests. Given that Fig. 4 shows strongly multimodal distributions of final fractions, mean values may not be representative, and differences between curves could be driven by rare replicates. The authors should report a measure of variability (e.g., error bars, interquartile ranges, or bootstrap confidence intervals) and perform appropriate statistical comparisons to support these comparisons.","section":"Fig. 2 and Section III"},{"comment":"The bridge-barrier mechanism is inferred from a single illustrative simulation run (Fig. 3) and from aggregate patterns (Figs. 4 and 5), but no quantitative evidence is provided that bridge areas actually slow or stop the spread of defection. For example, the authors could measure the time spent by the defector front at bridges, compare invasion speeds in assorted versus non-assorted networks, or test the fitness condition in Eq. (6) for nodes located at bridges. Without such quantitative analysis, the mechanism remains a post hoc interpretation rather than a demonstrated cause of the observed enhancement.","section":"Figs. 3-5 and Section IV"},{"comment":"All simulations use a single strategy-update rule: synchronous proportional imitation with the specific normalization in Eq. (3). This rule is known to interact strongly with degree heterogeneity, and it is plausible that the reported enhancement, especially the strong effect in degree-assorted Poisson networks, is specific to this update dynamics. To support the broad conclusion that correlations between cooperativeness and social connectedness affect the evolution of cooperation, the authors should test at least one alternative update rule (e.g., asynchronous updating, the Fermi rule, or birth-death dynamics) and show that the central results are robust.","section":"Section V.A, Eq. (3)"}],"minor_comments":[{"comment":"The capitalization in 'We use the following strategy assignment procedures' should be lowercase 'we' to be consistent with the surrounding text style.","section":"Section V.C"},{"comment":"The three curves in each panel are distinguished only by a legend; consider using different line styles or markers as well, so that the figure is readable in grayscale print.","section":"Fig. 2"},{"comment":"The distributions in Fig. 4 combine final fractions for all game parameter values into one histogram; separating panels by b or rho would be more informative, since the distinct peaks likely correspond to different parameter regimes.","section":"Fig. 4"},{"comment":"The sentence 'The inequality is always fulfilled for d = n'(1+rho)/(1-rho)' is confusing because the condition is stated as an equality; clarify that the inequality holds when d exceeds this threshold value.","section":"Appendix A, Eq. (7)"},{"comment":"Reference [32] is a note rather than a citation; it should be moved into the main text or a footnote.","section":"Reference [32]"},{"comment":"Please clarify whether the 50 replications for each setting use independent network realizations or the same network with different initial strategy assignments, as this affects the interpretation of the variability.","section":"Section II.D"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of q-bio.PE and addresses an interesting question. The stationarity issue is the main concern; if the authors can show the effect persists over longer horizons or provide a convincing argument that bridges are stable, the paper could be suitable. The lack of code/data is not a blocker for this journal, but would strengthen reproducibility. Please also ensure that the revised version addresses the lack of statistical support for the central comparisons."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The thing to know about this paper: it finds a genuine reversal that I have not seen stated before. In standard networks, well-connected cooperators help in scale-free but not Poisson networks. Once you add degree assortativity, the opposite can hold: Poisson networks with perfect strategy-degree correlation maintain cooperation surprisingly well, even exceeding scale-free, because bridge areas between degree clusters block defection. That is a real contribution to the network-cooperation literature, and it is framed honestly with both Prisoner's Dilemma and Snowdrift results, clear parameter choices, and 50 replications per setting. The Trojan-horse mechanism—defectors seeded inside cooperator clusters destroying the bridge protection—is also a nice, plausible explanation for why intermediate correlation fails where perfect correlation succeeds.\n\nThe soft spots are real but not fatal. The headline plots have no error bars or statistical tests, and the bridge-barrier mechanism is inferred from illustrative runs rather than quantitative measures of bridge structure. The bigger concern is the stationarity question. The paper reports the mean cooperator fraction over the last 100 of 10^4 timesteps, but never shows the system has settled. Its own Fig. 5 shows the defector fraction advancing in stair-step jumps, which looks like an ongoing slow invasion, not equilibrium. The text even concedes a defector cluster might eventually overcome a bridge. So the “maintains high levels above 20%” claim, and the comparison with scale-free networks, could erode if tmax were extended. That is a load-bearing uncertainty, not a minor quibble. Also, everything rests on one update rule—synchronous proportional imitation—which is known to interact strongly with degree heterogeneity.\n\nOn the citation pattern: the relevant prior work on initial distributions [20] and degree assortativity [18] is acknowledged, and the new twist is the systematic combination of strategy-degree correlation with assortativity plus the bridge mechanism. No code or data is shipped, which is a shame for a simulation paper.\n\nWho is this for? Anyone studying cooperation on networks, especially people designing human cooperation experiments where initial placement can matter. It deserves a serious referee. My recommendation: send it to review, but the revision must include stationarity checks, error bars, robustness to at least one alternative update rule, and ideally code release. The core idea is worth the effort.","headline":"A well-specified simulation study with a real and interesting reversal—degree-assorted Poisson networks can keep cooperation high when cooperators occupy hubs—but the persistence claim needs stationarity checks before it is fully load-bearing.","tokens_in":12997,"tokens_out":1122,"would_cite":true,"duration_ms":13119,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["91A22","91D30","92D15"],"pacs":[],"model":"deepseek-v4-flash","headline":"Degree-assorted Poisson networks keep cooperation alive at every temptation level when cooperators occupy the hubs.","keywords":["evolution of cooperation","degree assortativity","cooperator-degree correlation","Prisoner's Dilemma","Snowdrift game","proportional imitation","Poisson networks","scale-free networks"],"falsifier":"Run the same simulations on degree-assorted Poisson networks with perfect cooperator-degree correlation but asynchronous updating or a Fermi update rule, and check whether the average final cooperator fraction at $b=2$ drops below the paper's roughly 20 percent level; if it does, the barrier effect is an artefact of the specific update rule rather than a property of degree-assorted structure.","tokens_in":12071,"feed_emoji":"🤝","tokens_out":8173,"duration_ms":78326,"temperature":0.7,"pith_summary":"This paper asks whether cooperation survives longer when cooperative individuals occupy well-connected positions in a social network. The authors find that the answer depends on network structure: in standard scale-free networks, putting cooperators on high-degree nodes raises final cooperation, while in standard Poisson networks it does nothing. The central result is that when networks are rewired so that nodes of similar degree cluster together, the same perfect cooperator-degree correlation causes bridge areas between clusters to act as barriers to the spread of defectors, keeping average final cooperator fractions above about 20 percent in Poisson networks for all values of the Prisoner's Dilemma temptation parameter $b>1$ and even exceeding scale-free levels. This protection is fragile: defectors placed inside cooperator clusters act as Trojan horses that let defection invade. The findings matter because real social networks show degree assortativity, so chance correlations between cooperativeness and connectedness could shape both real cooperation and the outcome of network-based cooperation experiments.","feed_headline":"Well-connected cooperators hold cooperation above 20 percent","feed_subtitle":"In degree-assorted Poisson networks, bridge zones block defector spread and can even beat scale-free levels.","key_machinery":"The mechanism that carries the argument is the combination of degree-assorted network topology and a single strategy-update rule: synchronous proportional imitation, where a node copies a neighbour's strategy with probability proportional to the fitness difference divided by the larger of the two degrees (Eq. 3). Degree assortativity is created by a rewiring algorithm that repeatedly reconnects nodes of similar degree, producing clusters of degree-similar nodes connected by bridge areas. With perfect cooperation-degree correlation, cooperators sit in the high-degree clusters, and the bridges between clusters are the points where defectors' spread is held back. The fitness inequality behind this appears in Appendix A: in the Prisoner's Dilemma, a cooperator with $d$ neighbours of whom $n$ cooperate has higher fitness than a defector with $n'$ cooperator neighbours only if $n > b n'$, so a defector needs many cooperator neighbours to outcompete a cooperator. This makes the bridge nodes, which have few cooperator neighbours, poor places for defection to cross.","core_discovery":"The paper's central claim is that correlation between cooperativeness and node degree is a real evolutionary force, but only when the network's wiring lets that correlation translate into protected clusters. In standard networks, placing cooperators on high-degree nodes helps only in scale-free networks, where hubs can act as strong role models; in standard Poisson networks, the degree advantage of cooperators is too small to change final outcomes. In networks with increased degree assortativity, however, nodes of similar degree are connected to each other, and when cooperators deterministically occupy all high-degree nodes, the bridge areas between clusters become barriers that stop the defector strategy from spreading from one cluster to the next. The dynamics show defectors rapidly taking over a cluster, then stalling at the bridges, which leaves high final cooperator fractions and produces a characteristic step-like rise in the average degree of defectors. The same mechanism explains the fragility under intermediate correlation: stochastic placement seeds a few defectors inside cooperator clusters, and these Trojan horses let defection start from within, bypassing the barriers.","pith_inferences":["A natural test: replacing synchronous updating with asynchronous proportional imitation may weaken the bridge barriers, because a single defector at a bridge node could be copied before the cluster-wide fitness advantage reasserts itself; the paper reports results only for synchronous updating.","The Trojan-horse vulnerability suggests a sharper prediction than the paper states explicitly: deliberately placing one or two defectors on bridge nodes rather than randomly inside a cluster should collapse the barrier almost immediately, whereas the same number placed elsewhere would barely matter.","The same mechanism might be exploited in network design: if a planner controls initial placement, assigning cooperative individuals to the highest-degree nodes in an assortative network is a cheap way to keep cooperation alive, but only if no defectors are seeded inside the cooperator clusters.","Because real animal and human networks show both degree assortativity and long-term structural stability, a relevant empirical check is to measure whether cooperative individuals in real populations actually sit on higher-degree nodes; if they do, the model predicts bridge positions should show distinctive strategy-transition dynamics."],"forward_implications":["In degree-assorted Poisson networks with perfect cooperator-degree correlation, average final cooperation stays above roughly 20 percent for every temptation parameter $b>1$ in the Prisoner's Dilemma, and can exceed cooperation in scale-free networks.","Stochastic (intermediate) correlation between cooperativeness and degree removes most of this benefit when degree assortativity is high, because the few defectors seeded inside cooperator clusters act as Trojan horses.","The apparent rule that higher degree heterogeneity promotes cooperation is not universal; under degree assortativity and perfect correlation, Poisson (low-heterogeneity) networks can outperform scale-free ones.","The bridge-barrier dynamics produce multi-peaked distributions of final cooperator fractions and stepwise increases in defector average degree, giving signature statistics that can be checked in runs or experiments.","The results are relevant to the design of network cooperation experiments, since initial random placement of cooperators can create chance correlations with degree that materially shift outcomes in degree-assorted networks."],"supporting_citations":[{"why":"Supplies the proportional-imitation update rule and the baseline scale-free result that the paper extends.","marker":"[9]"},{"why":"Establishes the standard positive relation between degree heterogeneity and cooperation that the degree-assorted Poisson result overturns.","marker":"[16]"},{"why":"Reports that degree assortativity alone can reduce cooperation, the baseline against which the bridge-barrier enhancement is measured.","marker":"[18]"},{"why":"Provides the Snowdrift game parameterisation and a second source for the proportional imitation update rule.","marker":"[28]"},{"why":"Gives the rewiring algorithm used to create degree-assorted networks while preserving their degree distributions.","marker":"[31]"},{"why":"Defines the standard Poisson random network baseline used for the central comparison.","marker":"[29]"},{"why":"Defines the standard scale-free network baseline that the degree-assorted Poisson result exceeds.","marker":"[30]"}],"fun_headline_variants":["Bridge barriers block defector spread in assortative networks","Degree assortativity lets Poisson networks beat scale-free","Trojan horses in clusters let defection invade","Well-connected cooperators only help in scale-free networks","Bridge areas act as barriers, Trojan horses break them"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that individuals update strategies synchronously by proportional imitation with normalization by the larger degree; the claimed barrier effect may not survive other common update rules, which the paper does not test.","fun_headline_variants_meta":{"raw":{"variants":["Bridge barriers block defector spread in assortative networks","Degree assortativity lets Poisson networks beat scale-free","Trojan horses in clusters let defection invade","Well-connected cooperators only help in scale-free networks","Bridge areas act as barriers, Trojan horses break them"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00058,"raw_usage":{"total_tokens":2748,"prompt_tokens":974,"completion_tokens":1774,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":590,"completion_tokens_details":{"reasoning_tokens":1700}},"tokens_in":590,"tokens_out":1774,"duration_ms":13422,"temperature":1.0,"reasoning_tokens":1700,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:00:21.376291+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same simulations on degree-assorted Poisson networks with perfect cooperator-degree correlation but asynchronous updating or a Fermi update rule, and check whether the average final cooperator fraction at $b=2$ drops below the paper's roughly 20 percent level; if it does, the barrier effect is an artefact of the specific update rule rather than a property of degree-assorted structure.","supporting_citations":[{"cited_title":"Scale-free networks provide a unifying framework for the emergence of coop- eration,","cited_arxiv_id":null,"evidence_quote":"Supplies the proportional-imitation update rule and the baseline scale-free result that the paper extends."},{"cited_title":"A new route to the evolution of cooperation,","cited_arxiv_id":null,"evidence_quote":"Establishes the standard positive relation between degree heterogeneity and cooperation that the degree-assorted Poisson result overturns."},{"cited_title":"Roles of mixing patterns in cooperation on a scale-free networked game,","cited_arxiv_id":null,"evidence_quote":"Reports that degree assortativity alone can reduce cooperation, the baseline against which the bridge-barrier enhancement is measured."},{"cited_title":"Spatial struc- ture often inhibits the evolution of cooperation in the snowdrift game,","cited_arxiv_id":null,"evidence_quote":"Provides the Snowdrift game parameterisation and a second source for the proportional imitation update rule."},{"cited_title":"Reshuffling scale- free networks: From random to assortative,","cited_arxiv_id":null,"evidence_quote":"Gives the rewiring algorithm used to create degree-assorted networks while preserving their degree distributions."},{"cited_title":"On the evolution of ran- dom graphs,","cited_arxiv_id":null,"evidence_quote":"Defines the standard Poisson random network baseline used for the central comparison."},{"cited_title":"Emergence of scaling in random networks,","cited_arxiv_id":null,"evidence_quote":"Defines the standard scale-free network baseline that the degree-assorted Poisson result exceeds."}],"review_version":1}