{"id":"2f41ed73-1410-4e97-89e1-f60836ce77be","arxiv_id":"1908.06883","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":9,"one_line_summary":"A modified Sugarscape rebellion model links grievance to local resources and shows how police spending and poverty assistance change uprising patterns.","lead":"The paper adapts a well-known Netlogo rebellion simulation to a Sugarscape world with resources, taxes, and police funding, then watches unrest under different policies. It is useful as a demonstration that simple agent models can show how poverty relief and policing shape protests in an artificial society.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central case-study findings rest on single stochastic runs with no code, seeds, or sensitivity analysis, so the claimed emergent patterns are not yet established as robust model properties.","rationale":"The reader's weakest_assumption identifies the grievance-resource link as the load-bearing assumption, and that is indeed a substantive modeling choice. However, the paper explicitly states it is not validating the theory but applying it to an artificial society, so the grievance assumption is an acknowledged premise rather than an insecure evidence claim. The more directly load-bearing issue for the central claim is the absence of reproducibility and sensitivity analysis. The paper's evidence for emergent policy-relevant phenomena consists entirely of qualitative observations from stochastic simulations, with no code, seeds, repeated runs, or parameter sweeps. The keyword 'sensitivity analysis' makes this omission especially pointed. This concern aligns with the reader's rationale, which explicitly lists 'absence of code, data, repeated runs, and sensitivity analysis' as a main weakness, even though the formal weakest_assumption field focuses on the grievance mapping. Therefore I partially agree with the reader. The appropriate verdict remains CONDITIONAL: the model may be a useful demonstration, but the authors must release the artifact and demonstrate that the qualitative results are stable across seeds and parameter choices. My concern strengthens the case for that condition but does not change the verdict.","tokens_in":5908,"tokens_out":3223,"duration_ms":34953,"concrete_test":"Release the NetLogo code and run each case in Table 1 with at least 20 random seeds, reporting means and ranges for cop count, rebel count, and jailed count over time. Then sweep the two key policy parameters, wealth-to-create-cop and wealth-distribution, over {0, 0.25, 0.5, 0.75, 1.0} while holding other parameters fixed, and check whether the qualitative transitions (cop collapse, constant rebellion, pacification) persist across seeds and neighboring parameter values. If the patterns are not stable, the case-study conclusions do not support the central claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim is that the rebellion-on-sugarscape model 'can be used to investigate emergent phenomena and implications of governmental policies.' The evidence for this claim consists of qualitative case-study observations in Section 4, e.g., that spending all government wealth on new cops reduces the cop force to 90 and produces constant rebellion (Section 4.2), while redistributing half the wealth to the poor nearly eliminates rebellion. These are NetLogo agent-based simulations, which are stochastic by default, yet the paper reports no repeated runs, no random seed information, and no variance measures. The keyword list promises 'sensitivity analysis' but no sensitivity analysis appears anywhere in the text. Without such analysis, the reported differences (cop count 158 vs. 90, jailed 18-20 vs. 0-5) could be single-run noise or artifacts of the particular parameter values chosen. The model description also omits the exact functional form linking local sugar level to perceived hardship, and the precise rules for cop maintenance and death, so the simulation cannot be reproduced from the paper alone. Consequently, the case-study conclusions do not yet demonstrate robust model behavior, and the central claim about the model's utility for policy investigation is unsupported by the evidence presented.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents an agent-based model called 'rebellion on sugarscape,' which overlays Epstein and Axtell's sugarscape resource landscape onto Epstein's rebellion model. Each agent's grievance is determined by the sugar level of its location, the government collects taxes to maintain cops, create new cops, or assist the poor, and agents may be mobile or immobile. The paper reports qualitative case studies: a baseline case, a comparison of spending on new cops versus redistribution to the poor, and additional cases varying movement, population, tax rate, vision, and legitimacy. The headline findings are that spending all government wealth on new cops actually reduces the cop force and produces constant rebellion, while redistributing wealth to the poor nearly eliminates rebellion. The paper claims the model can be used to investigate emergent phenomena and implications of governmental policies.","tokens_in":6221,"tokens_out":3363,"duration_ms":39911,"significance":"If the model were fully specified and its qualitative conclusions supported by repeated runs and sensitivity analysis, this would be a useful pedagogical demonstration linking a spatial resource map to greed-and-grievance conflict dynamics and a plausible illustration that simplistic policy interventions can backfire. The paper is honest about not validating the underlying theory, and the idea of tying grievance to a geographic resource layer is a reasonable extension of the original rebellion model. The case-study structure is clear and policy-relevant. However, the current evidence is not sufficient to establish the claimed emergent phenomena: the observations rest on single stochastic runs, the model equations are not given, no code or seeds are provided, and the principal 'poor regions rebel' result is partly built into the design.","major_comments":[{"comment":"The observation that rebels concentrate in poor regions is not an emergent simulation finding; it is entailed by the model definition. Section 3.1 states that an agent in a sugar-rich region has low perceived hardship and vice versa, so higher-grievance agents are placed in low-sugar locations by construction. The baseline case in Section 4.1 therefore restates the design rather than validates the model. Similarly, the pacifying effect of poverty assistance follows directly because raising the wealth of poor agents lowers their grievance. The paper should either label these as illustrative consequences of the assumptions or run tests with alternative grievance formulations, such as relative wealth or inequality within a neighborhood, to demonstrate genuine emergence.","section":"§3.1 and §4.1"},{"comment":"The central 'surprise' — that spending all government wealth on new cops reduces the cop force to 90 and produces constant rebellion — is reported from a single simulation run. The explanation depends on several unstated details: the per-cop maintenance cost, the tax base, the timing of cop death versus new-cop creation, and the random placement of new cops. No repeated runs with different random seeds and no parameter sweeps are reported, so the qualitative inversion from 'spend on cops' to 'weaker policing' could be an artifact of the chosen parameter values or of stochastic variation. The paper needs at least a range of values for the maintenance cost, tax rate, and new-cop cost, plus variance information, before this case can be claimed as robust model behavior.","section":"§4.2"},{"comment":"The model is not reproducible from the manuscript. The exact functional forms for grievance, perceived hardship, legitimacy, risk, arrest probability, tax collection, cop maintenance, and cop death are never written down; Figure 1 shows only formula fragments and does not define all variables. No NetLogo code, parameter table, initial configuration, or random seed is provided. Without a complete specification, the reader cannot check that the reported cop counts and jail counts follow from the stated rules, and the claimed case-study phenomena cannot be independently reproduced.","section":"§3 and Figure 1"},{"comment":"The keyword list advertises 'sensitivity analysis,' but no sensitivity analysis appears anywhere in the text. Table 1 varies some parameters across cases, but each entry is a single unstructured run with no repeated-run statistics, error bars, or systematic one-factor-at-a-time exploration. Several rows report only qualitative outcomes such as 'No rebel' without quantitative support. Consequently, the paper's central claim that the model 'can be used to investigate emergent phenomena and implications of governmental policies' is not yet supported by the evidence presented.","section":"Keywords and §4"}],"minor_comments":[{"comment":"The diagram contains truncation-like formula fragments and undefined symbols; it should be redrawn with complete, labeled equations for each relationship shown.","section":"Figure 1"},{"comment":"There are citation inconsistencies: the text cites Collier and Hoeffler (2002) but the reference list gives the 2000 World Bank paper, and the Epstein (2002) reference is missing the correct venue and page details for the PNAS article.","section":"References"},{"comment":"The table would be more informative if each case included the number of runs, the variance of key outputs, and explicit numerical values for cop counts, rebel counts, and jail counts rather than qualitative descriptions.","section":"Table 1"},{"comment":"The phrase 'constant rebellions' is used to describe the all-cop-funding scenario, but no time-series plot or frequency measure is given; a temporal trace of active rebels would make the contrast with the baseline's periodic outbursts concrete.","section":"§4.2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript reads more like a work-in-progress workshop note than a finished journal article. If the journal values short exploratory ABM studies, major revision with repeated runs, a full model specification, and sensitivity analysis could bring it to publishable form; without those additions, the central claims remain unsupported."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nRead this one so you know what an ABM policy-demonstration paper looks like when it is honest but not yet rigorous. The author grafts a Sugarscape resource layer onto Epstein's NetLogo rebellion model, adds taxation, police budgets, and poverty transfers, then walks through a handful of case studies. The genuinely interesting bit is in Case 2: when the government pours all its wealth into creating new cops, the cop force collapses to 90 and rebellion becomes constant, while redistributing half the wealth to the poor nearly eliminates rebellion. That is a non-obvious trade-off worth explaining.\n\nWhat the paper does well: it is clearly written, the model rules are described plainly, and Section 5 does not oversell. The author correctly says the model is far from real-world motivations. The citation pattern is reasonable, with the relevant Epstein, Collier-Hoeffler, and critique literature. As an incremental extension of an existing model, it is a genuine contribution, just not a large one.\n\nThe soft spots are real and load-bearing. First, the key 'rebels concentrate in poor regions' observations are largely built into the model: grievance is defined as low local wealth in Section 3.1, so Section 4.1 is a restatement of the design. The pacifying effect of assistance follows similarly. Second, every case study is a single stochastic run. NetLogo is stochastic by default; the paper reports no seeds, no repeated runs, no variance, and no sensitivity analysis despite listing 'sensitivity analysis' as a keyword. The cop-force drop from 158 to 90 could be parameter noise or an artifact of the specific cost values. Third, the exact functional form linking sugar level to perceived hardship is not given, so the simulation is not reproducible from the text alone. These are not minor quibbles; they undermine the central claim that the model can be used to investigate emergent policy implications.\n\nMy take: the model is plausible and the author is thinking clearly, but the paper as written is not yet a robust scientific artifact. It reads like a workshop demonstration. I would not send it to peer review in its current form; I would encourage a resubmission that ships the NetLogo code, states seeds and parameters, runs a batch sensitivity analysis, and separates the by-construction results from the emergent surprises. If that were done, the cop-funding result might earn a real paper.","headline":"A clearly written but under-supported ABM extension; the one surprising cop-budget result needs code, seeds, and sensitivity analysis before it can be trusted.","tokens_in":6660,"tokens_out":2831,"would_cite":false,"duration_ms":29496,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that an agent-based model linking grievance to local resources can reveal emergent government-policy consequences, including the finding that spending all tax wealth on new police can shrink the effective force and…","keywords":["agent-based model","civil conflict","greed and grievance","sugarscape","rebellion model","NetLogo","public policy","emergent phenomena"],"falsifier":"Compare subnational conflict data with local resource wealth while holding police capacity fixed: if conflict events do not concentrate in the poorest regions, the model's core resource-to-grievance mapping is falsified.","tokens_in":5698,"feed_emoji":"👮","tokens_out":6257,"duration_ms":58802,"temperature":0.7,"pith_summary":"This paper builds an agent-based model of civil conflict by grafting the standard rebellion model onto a sugarscape, a terrain with unevenly distributed resources. Its central aim is to show that linking an agent's grievance to the resources at its location allows the model to produce credible qualitative policy lessons. Under the model, a government that pours all tax revenue into creating new police can end up with fewer effective cops and constant rebellion, while redistributing wealth to the poor nearly eliminates rebellion. The paper offers this as a demonstration that marrying civil-conflict theory with a resource map yields emergent phenomena worth studying, not as a validated empirical claim about real civil wars.","feed_headline":"Spending all tax money on new cops backfires","feed_subtitle":"Model of rebellion on a resource landscape finds police buildup shrinks the force, while helping the poor keeps peace.","key_machinery":"The mechanism is the resource-linked grievance function on a sugarscape. Each agent's grievance is $G=H(1-L)$, where $H$ is perceived hardship set by the sugar level at the agent's location and $L$ is perceived government legitimacy, so poor regions systematically breed rebels. The government collects a tax on agents' sugar income and divides it among maintaining cops, creating new cops (cost 10 sugars apiece), and assisting people below the poverty line of 1 sugar. This feedback loop—local resources determine grievance, grievance determines rebellion, rebellion and jailing determine tax revenue, tax revenue determines police and transfers—carries all the case-study results.","core_discovery":"On the paper's own terms, the central discovery is that policy choices interact with resource geography to produce surprising macro-outcomes from simple micro-rules. When the government spends all its tax revenue on building new cops rather than maintaining them, the cop force falls from 100 to about 90 and a constant mass of rebels appears; when it instead redistributes half its wealth to people below the poverty line, rebellion almost disappears while the original 100 cops are preserved. These outcomes follow from the model's mechanism: grievance is high where sugar is low, jailed rebels stop paying taxes, new cops are expensive and randomly placed, and poverty assistance lowers grievance directly. The paper also reports that with freedom to move, rebellion shifts to the boundary between rich and poor regions, and that government legitimacy below 0.75 makes the government collapse.","pith_inferences":["Read as a model of fiscal feedback, the most counterintuitive result implies that police expansion can be self-defeating: jailing rebels cuts tax revenue, starving the very force meant to grow; a real-world test would compare police funding levels against conflict or crime data with a time lag.","The model predicts that, holding police capacity fixed, conflict events should concentrate in resource-poor regions; this is testable with subnational data on conflict, wealth, and government transfers.","A natural extension would let grievance depend on relative wealth inside an agent's neighborhood rather than absolute local sugar; if that variant erases the poverty-rebellion link, the policy lesson shifts from transfers to inequality reduction."],"forward_implications":["If the government spends all collected wealth on new cops, the effective police force falls and rebellion becomes constant rather than periodic.","Wealth redistribution to the poor keeps rebellion near zero while preserving the existing police force.","Allowing movement shifts rebellion to the boundary between rich and poor regions, not into the richest areas.","Government legitimacy at or below about 0.75 leads to government collapse and more rebels than quiet citizens.","A high tax rate of 50 percent produces mass rebellion in poor regions even when some wealth is redistributed."],"supporting_citations":[{"why":"Supplies the greed-and-grievance theory of civil war that the model operationalizes.","marker":"Collier and Hoeffler (2002)"},{"why":"The original rebellion model that the paper extends by adding a resource layer and government budget choices.","marker":"Epstein (2002)"},{"why":"Provides the sugarscape artificial-society environment used for the resource map.","marker":"Epstein and Axtell (1996)"},{"why":"The NetLogo rebellion model that the simulation code is based on.","marker":"U. Wilensky (2004)"},{"why":"The feasibility hypothesis used to frame why opportunity and motivation both matter for civil war.","marker":"Collier, Hoeffler and Rohner (2008)"}],"fun_headline_variants":["Cops vs. cash: why police buildup fuels rebellion","Policing backfires: model shows helping poor quells unrest","Taxing to build cops shrinks police force, sparks rebels","On Sugarscape, redistribution beats police buildup","Spend on poor, not police: model's recipe for peace"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that an agent's grievance is set by the sugar level at the place it lives, so people in resource-poor regions are automatically more aggrieved and assistance to them automatically lessens grievance; if real grievances track relative deprivation, inequality, or non-economic factors, the model's lessons do not transfer.","fun_headline_variants_meta":{"raw":{"variants":["Cops vs. cash: why police buildup fuels rebellion","Policing backfires: model shows helping poor quells unrest","Taxing to build cops shrinks police force, sparks rebels","On Sugarscape, redistribution beats police buildup","Spend on poor, not police: model's recipe for peace"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001026,"raw_usage":{"total_tokens":4248,"prompt_tokens":792,"completion_tokens":3456,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":408,"completion_tokens_details":{"reasoning_tokens":3373}},"tokens_in":408,"tokens_out":3456,"duration_ms":24891,"temperature":1.0,"reasoning_tokens":3373,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:43:00.843423+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compare subnational conflict data with local resource wealth while holding police capacity fixed: if conflict events do not concentrate in the poorest regions, the model's core resource-to-grievance mapping is falsified.","supporting_citations":[],"review_version":1}