{"id":"d7f09d2e-d908-4318-af16-5423897ccb44","arxiv_id":"1908.02371","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"A simulation model shows that ecosystem engineering, particularly redundant engineering, stabilizes community assembly and increases diversity by lowering colonization barriers.","lead":"This paper adds ecosystem engineers to a network model of how ecological communities assemble, letting species modify the environment rather than only eat or help each other. In simulations, engineering and especially redundant engineering can make communities more diverse and more persistent, but low levels of engineering can raise extinction rates.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The unique-modifier control may not isolate redundancy: if total modifier count or engineer count differs between the redundant and unique conditions, the reported drop in Su* could reflect changed obligate dependencies rather than redundancy per se.","rationale":"The reader's conditional verdict is appropriate, but the most load-bearing weakness is not Eq. 1. The competition-strength ordering cn>ce>cv affects the mutualism, nestedness, and extinction-cascade narratives; however, the paper's central claim about redundancy is governed by the colonization rules and by the comparison between redundant and unique modifier assignments. The unique-modifier control is the crux of that comparison, and the manuscript does not specify whether the total number of modifier nodes or engineer species is held fixed. The SI derivation of E{M_P} makes the distinction concrete: if uniqueness is implemented by giving each maker-modifier pair a distinct identity, the expected number of modifier nodes rises from 0.632 η S to η S, which directly increases the expected number of obligate service dependencies. The observed drop in Su* could then be driven by a larger dependency load rather than by the absence of redundancy. The unstated modifier decay rate rm compounds this uncertainty, since the benefit of redundancy depends on how long unique modifiers survive after their engineers disappear. These issues do not disprove the redundancy claim; they make it impossible to verify from the present text. The concrete test would settle the matter by holding M and engineer count fixed across treatments and by scanning rm. Because the paper is otherwise carefully argued and the concern is a reproducibility/control-specification issue rather than a demonstrated contradiction, the verdict should remain conditional rather than move to acceptance or rejection.","tokens_in":16762,"tokens_out":13194,"duration_ms":159818,"concrete_test":"Obtain or reimplement the ENIgMa source-pool algorithm and run the unique-modifier control in three variants: (a) fix the realized number of modifier nodes to the redundant case and prune redundant makers, preserving the engineer set; (b) assign each maker-modifier pair a distinct modifier identity, allowing M to increase; (c) fix both M and the number of engineers while randomizing which single species makes each modifier. If Su*/S* < 1 occurs only under variant (b), the decline is an artifact of increased modifier count rather than redundancy. Independently, vary rm from 0.01 to 10 with rc=re=1 and η=1; if the redundant-versus-unique richness gap shrinks as rm approaches 0, the headline result depends on an unreported modifier persistence timescale.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that redundant engineering raises community diversity by lowering colonization barriers rests on comparing the standard ENIgMa model with a control where 'each modifier is uniquely produced by a single species' (main text, Fig. 4D; SI Figs. S6, S7). The paper's source-pool construction (SI Section I, Eq. S1) gives the expected number of modifier nodes as E{M_P}=S_P η(1−1/e), specifically because multiple species can independently be assigned the same modifier identity. The unique-modifier control is described only as 'the same model, but where each modifier is uniquely produced by a single species.' Two different implementations are possible: (i) keep the same total number of modifier nodes and the same engineer set, pruning all but one maker per modifier; or (ii) give every maker-modifier pair a distinct modifier identity, which changes the expected number of modifier nodes from S_P η(1−1/e) to S_P η. If implementation (ii) was used, then because each pair (species, modifier) has probability qn of being a need interaction, the unique condition has roughly 58% more modifier nodes and therefore more obligate service dependencies per species. That alone would make colonization harder and reduce steady-state richness, independent of whether redundancy matters. The paper does not specify which implementation was used. A second linked omission is the modifier decay rate rm: a modifier persists after its last maker goes extinct at rate rm, but rm is never reported. If rm is very small, unique modifiers remain available long after their engineer disappears, weakening the redundancy effect; if rm is large, the effect is amplified. The reported Su*/S* < 1 pattern is thus only interpretable relative to an unspecified control construction and an unreported timescale.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces ENIgMa, a stochastic presence/absence model of community assembly in which species interact through trophic (eat), service (need), and engineering (make) interactions, the last operating through abiotic modifier nodes. Colonization requires one eat interaction and all need interactions to be satisfied; primary extinction is driven by competitive exclusion parameterized by a linear competition-strength function, and secondary extinctions follow the loss of required interactions. Using Gillespie simulations with default parameters, the authors report that assembly without engineering reproduces several empirical patterns: generalists appear early under a steady-state-scaled generality measure, trophic-level distributions are hourglass-shaped, and networks become more nested as the frequency of mutualisms increases, at the cost of higher secondary extinction. Adding ecosystem engineers produces a nonmonotonic response of primary extinction rates, with rare engineering increasing primary extinction and abundant engineering suppressing it, and reduces secondary cascades. The paper's final claim is that redundancy in engineered effects (multiple species producing the same modifier) lowers colonization barriers and increases steady-state diversity relative to a control in which each modifier is unique to a single species.","tokens_in":17036,"tokens_out":5443,"duration_ms":58054,"significance":"If the main claims survive scrutiny, the paper provides a useful mechanistic framework for including ecosystem engineering in network-based assembly models and makes a specific, testable prediction: redundant engineering stabilizes diverse communities by reducing priority effects. The model is explicitly defined, simulation methods are standard and repeatable in principle, and the SI contains analytical derivations for expected modifier counts (Eqs. S1-S4), which are a genuine strength. However, the paper's qualitative conclusions rest on an uncalibrated ordering of competition coefficients and on a control condition for redundancy that is underspecified; both need to be resolved before the results can be taken as established.","major_comments":[{"comment":"The competition-strength function sigma_i = cn*n_i - ce*e_i - cv*v_i with the chosen coefficients cn=pi, ce=sqrt(2), cv=1 encodes the assumption that each additional mutualistic need increases competitive strength more than each added prey or predator decreases it. All of the qualitative results on nestedness under mutualisms (Fig. 3) and on the nonlinear effect of engineering on primary extinction rates (Fig. 4A) depend on this ordering, yet the paper does not provide any sensitivity analysis, empirical calibration, or biological justification for the ordering. If the coefficients were varied, for example if the vulnerability penalty exceeded the mutualistic benefit, the reported patterns could reverse. I ask the authors to either justify the ordering from data or first principles, or to systematically explore the coefficient space and show that the qualitative conclusions are robust over a nontrivial region of it.","section":"Materials and Methods, Eq. (1)"},{"comment":"The central claim that redundant engineering promotes diversity is based on comparing the ENIgMa model to a control 'where each modifier is uniquely produced by a single species.' The manuscript does not specify whether this control (i) keeps the same modifier set and engineer set and prunes all but one maker per modifier, or (ii) assigns each maker-modifier pair a distinct modifier identity. Under implementation (ii), the expected number of modifier nodes changes from S_P*eta*(1 - 1/e) (SI Eq. S1) to S_P*eta, an increase of roughly 58%; because all need interactions are obligate for colonization, the unique condition would then have substantially more mandatory dependencies, and the observed decline in Su* could be caused by these added dependencies rather than by the absence of redundancy. The authors must specify the implementation, and if (ii) was used, repeat the comparison while holding the number of modifier nodes or the number of need interactions per species fixed.","section":"Fig. 4D and SI Figs. S6-S7"},{"comment":"The parameter rm, the rate at which a modifier disappears after its last maker goes extinct, is defined but never given a numerical value in the text or SI. The persistence of modifiers after engineer extinction is essential to the proposed mechanism by which redundancy lowers colonization barriers, and the effect of the engineering timescale cannot be assessed without this value. Please report the default value of rm and, ideally, test the sensitivity of the redundancy result to rm, including the limit rm -> infinity where modifiers disappear instantly.","section":"Materials and Methods, 'A modifier is present'"},{"comment":"The main text states that generalists dominate early in assembly and specialists increase later (Fig. 2B), but SI Section III reports that this pattern is an artifact of scaling generality to the steady-state value of L*/S*. With the two alternative scalings shown in Fig. S4 (Gall_i and Ghetero_i), specialists actually dominate early in assembly, so the main-text claim depends entirely on the chosen normalization. Because reproducing the empirically observed role of generalists is one of the paper's headline results, the authors should either frame the result as scaling-dependent in the main text or provide an ecological argument for why the steady-state scaling is the appropriate one.","section":"Fig. 2B and SI Section III / Fig. S4"}],"minor_comments":[{"comment":"The sentence 'where nodes represent ecological entities such as populations of species and or the presence of abiotic modifiers affecting species such as (examples)' contains a placeholder phrase and an incomplete list; please replace it with concrete examples or remove the parenthetical.","section":"Materials and Methods"},{"comment":"The text refers to 'Pilai et al.' when citing the metacommunity assembly model, but the reference list gives 'Pillai et al.'; please correct the in-text citation and verify all author name spellings.","section":"Materials and Methods and reference list"},{"comment":"The default parameter list at the end of Materials and Methods gives S, pe, cn, ce, cv, and the number of iterations, but the values of pn, qe, qn, and rm, all needed to reproduce the simulations, are not reported explicitly; please provide a complete parameter table in the main text or SI.","section":"Materials and Methods, simulation parameters"},{"comment":"The x-axis of Figs. 3 and 4 is labeled 'Frequency of service interactions' but the values are not defined in the figure captions; the relationship between this axis and pn should be stated.","section":"Figures 3 and 4"}],"recommendation":"major_revision","confidential_remarks":"The manuscript appears to be a preprint that has not been fully copyedited, as evidenced by the placeholder phrase in Materials and Methods. The most serious issue is the underspecified control for the redundancy claim: if the unique-modifier control was implemented by creating a separate modifier for every maker-modifier pair, then the main claim of the paper may be an artifact of changes in obligate dependency counts rather than of redundancy itself. I recommend requiring the authors to provide the exact implementation details and, if possible, the simulation code or a detailed pseudocode, as part of the revision. The paper is otherwise publishable in principle after the sensitivity and control analyses are addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: the paper extends the Pillai et al. assembly model with abiotic modifier nodes and 'make' interactions, and it finds a plausible, nontrivial result that redundant engineering lowers colonization barriers and raises steady-state richness. The model is clearly described, and the analytical expectation for redundant modifiers (Eq. S1) is a nice touch. The paper is honest about the generalist result depending on normalization (SI Section III). It is a legitimate within-subfield contribution.\n\nThe three soft spots, in order of severity. First, the unique-modifier control in Fig. 4D/S6/S7 is under-specified. The text says 'the same model, but where each modifier is uniquely produced by a single species,' but it does not say whether total modifier count is held fixed or whether every maker-modifier pair gets a distinct identity. If the latter, the unique condition has 1/(1−1/e) ≈ 1.58 times more modifier nodes than the redundant condition, and therefore more obligate need dependencies per species. More needs alone would suppress colonization and lower Su*/S*, so the reported drop does not cleanly isolate redundancy. The authors need to specify the construction and ideally run a matched control that keeps the number of modifier nodes and the number of engineer–modifier assignments constant. Second, the competition coefficients cn>ce>cv in Eq. 1 are load-bearing for the nestedness and primary-extinction results, but they are chosen ad hoc with no sensitivity analysis. A different ordering could reverse the patterns. Third, the paper does not report the event rates rc, re, rm, and no code or data are provided; rm especially matters because it controls how long a modifier lingers after its maker goes extinct, which directly bears on the redundancy mechanism. There is also a small internal inconsistency: the main text says the number of modifiers per species is drawn from Poiss(μ) with μ = ηe/(e−1), but Eq. S1 requires μ = η to give E{M_P}=Sη(1−1/e). Likely a typo, but it should be corrected.\n\nWho this is for: community ecologists and network modelers working on assembly, niche construction, and mutualism. It would be a reasonable paper after major revision. I would send it to review, and in the review I would insist on the control construction being clarified and sensitivity to cn, ce, cv being reported.","headline":"A useful model extension with a redundancy result that is likely real but currently rests on an under-specified control and an uncalibrated competition ordering.","tokens_in":17673,"tokens_out":4578,"would_cite":false,"duration_ms":44734,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Adding engineers to an assembly model, this paper claims that redundant engineering stabilizes communities by lowering colonization barriers and raising diversity, while sparse engineering raises primary extinction rates.","keywords":["ecosystem engineering","community assembly","ecological networks","nestedness","mutualism","species persistence","extinction cascades","niche construction"],"falsifier":"Re-run the simulations with the competition coefficients perturbed—for example with $c_v > c_e > c_n$ or with $c_n$ below $c_e$—and check whether the three headline results persist: increasing mutualism frequency still yields more nested networks, rare engineering still elevates primary extinction, and redundant engineering still raises steady-state richness. The claim would be falsified as stated if any of these patterns flips or disappears under a perturbed ordering, since the paper offers no other mechanism that would preserve them; a complementary empirical check would compare colonization success in real patches whose habitat modification is supplied by one engineer versus several redundant ones.","tokens_in":16537,"feed_emoji":"🕸️","tokens_out":13422,"duration_ms":118029,"temperature":0.7,"pith_summary":"The paper extends stochastic network models of community assembly to include ecosystem engineers: species that alter the environment in lasting ways, represented as additional 'modifier' nodes linked to species by make, eat, and need interactions. It claims that this model reproduces observed assembly features—generalists colonizing before specialists, hourglass trophic-level distributions, and increasing nestedness with mutualism frequency—and then identifies what engineering adds. The central claim is that ecosystem engineering has nonlinear effects on extinction—rare engineering raises primary extinction rates while abundant engineering suppresses both primary and secondary extinctions—and that redundancy in engineered effects is what stabilizes assembly, because multiple species producing the same modifier lowers the barriers to colonization and raises steady-state species richness. A sympathetic reader would care because the result offers a concrete mechanism for why functionally redundant habitat-modifying species may be what maintains biodiversity in real communities.","feed_headline":"Redundant ecosystem engineers stabilize community assembly","feed_subtitle":"Model: species sharing one habitat modification lower colonization barriers and raise species richness.","key_machinery":"The central object is the ENIgMa network model, in which nodes are species plus abiotic modifier nodes, and directed links are trophic ('eat'), service ('need'), or engineering ('make') dependencies; community dynamics are simulated as presence–absence events with a Gillespie algorithm. The load-bearing mechanism is the competition-strength function $\\sigma_i = c_n n_i - c_e e_i - c_v v_i$ with coefficients $c_n = \\pi > c_e = \\sqrt{2} > c_v = 1$, which encodes that each added mutualistic need increases competitive strength more than each added prey or predator decreases it; primary extinction hits any species that is not the strongest competitor for at least one of its resources. Two further rules do the specific work of the engineering result: modifiers persist stochastically after all their producers are gone, and each modifier is independently assigned to engineers, so that a fixed proportion of modifiers (about 42 percent) ends up produced by more than one species—this shared-production redundancy is what lowers colonization barriers and stabilizes diversity.","core_discovery":"In the ENIgMa assembly model (Eat–Need–Ignore–Make), where species and the abiotic modifiers they produce are present or absent and dynamics are driven by stochastic colonization and extinction events, the paper shows that ecosystem engineering alters assembly outcomes in ways that depend sharply on how common and how redundant engineering is. When each engineer produces a unique modifier, steady-state richness declines steeply because a species that needs several modifiers can only colonize after every required producer has arrived, magnifying priority effects. When multiple engineers can produce the same modifier, redundancy expands the accessible niche space, minimizes priority effects, and lowers the barriers to colonization, so a larger fraction of the source pool establishes and community diversity is maintained. The same model yields two further results: mutualism frequency increases nestedness at the cost of persistence, and a low density of engineers elevates primary extinctions of prey while compressing cascades, whereas a high density of engineers expands niche space and suppresses both primary and secondary extinction.","pith_inferences":["A conservation corollary the paper leaves implicit: protecting functionally redundant guilds of habitat modifiers matters for maintaining diversity even when no single engineer species is indispensable, because the redundancy itself, not any particular engineer, is what lowers colonization barriers.","Because modifiers persist after their makers vanish, the model implies that legacy engineering effects—beaver ponds, shell beds, soil structure—should buffer communities against extinction cascades on timescales longer than an engineer's lifetime; the paper includes this persistence but does not explore explicit time lags.","The analytically fixed redundancy fraction $\\varphi \\approx 0.418$, which is independent of $\\eta$, is a precise quantitative prediction of the random-assignment scheme that could be tested in real engineered systems by measuring how many environmental modifications are shared by multiple species.","An empirical test of the central mechanism: in habitat patches where a modification (burrows, dams, nests) is provided by several engineer species, incoming species needing that modification should show higher colonization success than in patches where a single engineer provides it."],"forward_implications":["Communities where several species produce the same environmental modification should reach higher steady-state species richness than otherwise identical communities with one-to-one engineering, because redundant modifiers admit more potential colonizers from the source pool.","Higher frequencies of mutualistic (service) interactions make assembled networks more nested but lower average species persistence by increasing secondary extinction rates, implying higher species turnover in mutualism-rich systems.","Sparse engineering ($\\eta \\leq 0.5$) raises primary extinction rates by stabilizing consumers and thereby increasing prey vulnerability, while abundant engineering ($\\eta > 0.5$) expands niche space and suppresses both primary and secondary extinction.","Even a small number of engineers reduces the magnitude of secondary extinction cascades, so disturbances in engineered communities tend to be compartmentalized into many small extinctions rather than a few large ones.","Network assembly models that omit engineering dependencies will systematically underestimate both a source of robustness (niche-space creation) and a source of fragility (larger cascades), and will miss the diversity-maintaining role of redundant engineers."],"supporting_citations":[{"why":"Supplies the presence/absence metacommunity framework whose colonization and extinction rules the ENIgMa model adapts.","marker":"[31]"},{"why":"Provides the trophic theory of island biogeography that motivates the colonization rules and the prediction that generalists arrive before specialists.","marker":"[33]"},{"why":"Offers the conceptual niche-construction framework with abiotic compartments that the modifier nodes of the model implement.","marker":"[13]"},{"why":"Establishes that engineering effects can outlast the engineer, justifying the persistence of modifier nodes after their producers vanish.","marker":"[23]"},{"why":"Documents the empirical nestedness of mutualistic networks that the model reproduces when service interactions are frequent.","marker":"[32]"},{"why":"Supplies empirical assembling food webs whose generalist-then-specialist pattern and connectance decay are compared with model output.","marker":"[34]"},{"why":"Provides the Niche model used as the structural baseline for comparing assembled network degree distributions.","marker":"[35]"}],"fun_headline_variants":["Redundant engineering lowers colonization barriers","Ecological engineering redundancy enhances community persistence","Redundant engineers expand niche space and boost diversity","Habitat modification sharing stabilizes ecological networks"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the coefficient ordering in the competition-strength equation $\\sigma_i = c_n n_i - c_e e_i - c_v v_i$, namely $c_n = \\pi > c_e = \\sqrt{2} > c_v = 1$, which asserts that each additional mutualistic need strengthens a competitor more than each additional prey or predator weakens it; the paper provides no sensitivity analysis or empirical calibration for this ordering, and if the ordering were reversed the reported patterns could reverse as well.","fun_headline_variants_meta":{"raw":{"variants":["Redundant engineering lowers colonization barriers","Ecological engineering redundancy enhances community persistence","Redundant engineers expand niche space and boost diversity","Habitat modification sharing stabilizes ecological networks"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000294,"raw_usage":{"total_tokens":1691,"prompt_tokens":904,"completion_tokens":787,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":520,"completion_tokens_details":{"reasoning_tokens":746}},"tokens_in":520,"tokens_out":787,"duration_ms":8649,"temperature":1.0,"reasoning_tokens":746,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:47:04.832540+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the simulations with the competition coefficients perturbed—for example with $c_v > c_e > c_n$ or with $c_n$ below $c_e$—and check whether the three headline results persist: increasing mutualism frequency still yields more nested networks, rare engineering still elevates primary extinction, and redundant engineering still raises steady-state richness. The claim would be falsified as stated if any of these patterns flips or disappears under a perturbed ordering, since the paper offers no other mechanism that would preserve them; a complementary empirical check would compare colonization success in real patches whose habitat modification is supplied by one engineer versus several redundant ones.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the presence/absence metacommunity framework whose colonization and extinction rules the ENIgMa model adapts."},{"cited_title":"Proc Natl Acad Sci USA 108(48):19293–19298","cited_arxiv_id":null,"evidence_quote":"Provides the trophic theory of island biogeography that motivates the colonization rules and the prediction that generalists arrive before specialists."},{"cited_title":"Nature Ecol Evol 1:0101","cited_arxiv_id":null,"evidence_quote":"Offers the conceptual niche-construction framework with abiotic compartments that the modifier nodes of the model implement."},{"cited_title":"BioScience 56(3):237–246","cited_arxiv_id":null,"evidence_quote":"Establishes that engineering effects can outlast the engineer, justifying the persistence of modifier nodes after their producers vanish."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents the empirical nestedness of mutualistic networks that the model reproduces when service interactions are frequent."},{"cited_title":"Proc Natl Acad Sci USA 100(16):9383–9387","cited_arxiv_id":null,"evidence_quote":"Supplies empirical assembling food webs whose generalist-then-specialist pattern and connectance decay are compared with model output."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Niche model used as the structural baseline for comparing assembled network degree distributions."}],"review_version":1}