{"id":"014578e4-263b-42c5-ae97-6d4d50c9a72c","arxiv_id":"1908.07974","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":2.0,"correctness_risk":"low","formal_verification":"none","parameter_count":5,"one_line_summary":"The authors re-derive optimal control conditions for existing Ebola SIR/SEIR models using Pontryagin's maximum principle and show numerically that these controls reduce simulated infections.","lead":"This paper applies standard optimal control theory to SIR and SEIR models of Ebola, adding vaccination, treatment, and education as time-dependent controls. A generalist might read it as a worked example of how epidemic control strategies are formulated and simulated.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"SIR Strategy 1 numerical results omit the cost weight B; without B, and with no code supplied, the figures underpinning the main conclusion are not reproducible.","rationale":"I checked the optimality systems in Sections 2 and 3 against the stated Hamiltonians. The adjoint equations and control projections in Theorems 1-5 are algebraically correct: the signs and the cost-weight denominators match the standard Pontryagin calculation, and the box constraints are applied properly. Thus the mathematical derivation is sound. The load-bearing weakness is in the numerical evidence for the stated conclusion. The paper's strongest claim is explicitly drawn from numerical illustrations, and those illustrations for SIR Strategy 1 depend on the unreported weight B in equation (4). Section 2.3 assigns D, C1, C2, and C3, but not B, and D is not the SIR Strategy 1 weight. This is compounded by the absence of code and of any convergence criterion, so the figures cannot be reproduced from the text alone. The reader's weakest assumption focused on population structure and parameter provenance; my concern is adjacent but more specific: even within the stated model, the numerical results are not fully specified. The reader's conditional verdict remains appropriate because the derivations are recoverable and the missing information could be supplied; the concern does not invalidate the central mathematics but it does make the numerical support conditional on an omitted input.","tokens_in":12851,"tokens_out":9907,"duration_ms":108406,"concrete_test":"Re-implement the forward-backward sweep for SIR Strategy 1 with ν=0.2, δ=0.1, s0=0.95, i0=0.05, r0=0 for B∈{0.1,1,5,20}, and compare the infected trajectories and η(t) with Figures 2 and 4. If all B values produce materially similar qualitative reductions and B=1 reproduces the figures, the omission is cosmetic; if the curves differ substantially across B, the reported numerical conclusion is quantitatively under-determined by the paper.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline conclusion is explicitly a numerical one: the paper claims that its illustrations show vaccination, education, and treatment reduce transmission. For the SIR model, that evidence depends on Strategy 1, which minimizes J(η)=∫[i(t)+(B/2)η²(t)]dt in equation (4), with optimal control η* given by equation (9). Section 2.3, however, never assigns a value to B; it lists D=1, C1=1, C2=5, C3=5, but D is the SEIR vaccination weight, not the SIR Strategy 1 weight. The trajectory in Figure 2 and the control in Figure 4 for Strategy 1 therefore depend on an unreported parameter. No code, no convergence criterion for the forward-backward sweep, and no final state or objective values are given, so the numerical support cannot be independently checked. This is not a mathematical invalidation: the adjoint equations and control characterizations are algebraically correct. The issue is that the central claim's evidence base is under-specified. If B was intended to be 1, a one-line statement would close the gap; as written, the numerical conclusion is conditional on an unknown input.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper applies Pontryagin's maximum principle to deterministic SIR and SEIR compartmental models of Ebola with time-dependent controls (vaccination, treatment, and educational campaign). The authors derive Hamiltonian functions, adjoint systems, and optimal control characterizations for several control strategies, including what they present as an additional SIR strategy with treatment and education controls, and they solve the optimality systems numerically using a forward-backward sweep with fourth-order Runge-Kutta. The central claim, based on numerical illustrations, is that all control strategies reduce the infected, exposed, and susceptible proportions relative to the uncontrolled model.","tokens_in":13104,"tokens_out":5526,"duration_ms":49775,"significance":"The algebraic derivation of the adjoint equations and optimality conditions is, for the stated Hamiltonians, standard and correct in the main cases checked, and the forward-backward sweep scheme is the recognized method for such problems. The paper, however, offers no existence or uniqueness theorems for the optimal controls, omits proofs for Theorems 2-5, and provides no code or complete parameter specification for the simulations, so the numerical evidence for the headline conclusion cannot be independently verified. The conclusion that controls reduce infection is also largely built into the model structure because vaccination/education remove susceptibles and treatment removes infecteds. As a purely methodological exercise applied to existing models, the work is incremental; its value depends on the completeness and reproducibility of the numerics.","major_comments":[{"comment":"The value of the weight B in the SIR Strategy 1 objective functional (4) is never assigned. Section 2.3 lists D=1, C1=1, C2=5, and C3=5, but D is the SEIR vaccination weight from Section 3.1, not B. Consequently the infected-proportion trajectory in Figure 2 and the control in Figure 4 for Strategy 1 depend on an unreported parameter, and the central numerical conclusion for the SIR model is not reproducible. A one-line statement of B's value (if B=1 was intended) would close this gap, but as written the evidence base is under-specified.","section":"§2.3, Eq. (4), Fig. 4"},{"comment":"The second equation of the controlled SIR system reads di/dt = ν s(t) s(t) − δ i(t); the correct term is ν s(t) i(t). If the numerical simulations were run with this erroneous state equation, the trajectories shown in Figures 1-4 would not correspond to the Hamiltonian in Eq. (6) or the adjoint system in Eq. (8). The equation must be corrected and the numerical implementation checked against the corrected system.","section":"§2.1, Eq. (5)"},{"comment":"Proofs of Theorems 2-5 are omitted, with only the statement that they are similar to Theorem 1. Since these theorems are the central analytical results of the paper, and the proof of Theorem 1 is itself only a sketch (it refers to solving ∂H/∂η=0 without showing the projection onto the box constraints), the paper should provide full proofs or at least a detailed derivation for one representative SEIR case, with the remaining cases relegated to an appendix.","section":"Remark 1 (after Theorem 2)"},{"comment":"The comparison of controlled trajectories against the uncontrolled model is not informative about the worth of the controls, because by construction the controls transfer susceptibles (vaccination/education) and infecteds (treatment) directly into the recovered class; any positive control therefore reduces the infected proportion. The conclusion would be strengthened by comparing strategies against one another (e.g., total cost, peak infection, or final size per unit cost) and explicitly stating that the paper only demonstrates that the modeled mechanisms do what they are built to do. This does not invalidate the mathematics but should be acknowledged in the interpretation.","section":"§2.3 and §3.4, no-control baseline"}],"minor_comments":[{"comment":"The text states that 'the positive parameter A is the weight on cost of vaccination,' but the objective functional uses B; no parameter A is defined.","section":"§2.1, after Eq. (5)"},{"comment":"In the model description, the compartments are listed as 'S(t), E(t), E(t) and R(t)', which should be 'S(t), E(t), I(t) and R(t)'.","section":"§3, Eq. (19)"},{"comment":"The fourth state equation is written as dr(r)/dt and should be dr(t)/dt.","section":"§3, Eq. (21)"},{"comment":"The objective functional and Hamiltonian contain mismatched brackets: J(η1,η2)=∫[D1 i(t)] + ... and H = D1 i(t)] + ...; the stray bracket after i(t) should be removed.","section":"§3.3, Eqs. (34) and (36)"},{"comment":"Theorem 4 states that the control minimizes 'over G3', but the control set defined for this strategy is G4; the reference should be updated.","section":"§3.2, Theorem 4"},{"comment":"In the Hamiltonian, the term 'δi + η(t)s(t)' should read 'δ i(t) + η(t) s(t)'.","section":"§3.2, Eq. (30)"}],"recommendation":"major_revision","confidential_remarks":"The paper is essentially a textbook-level application of standard optimal control theory to models from Rachah and Torres; the novelty is a single added strategy for the SIR model, and the numerical scheme is standard. For this journal, the contribution is quite incremental. However, if the authors can supply complete reproducible numerical details, correct the state-equation typo, and provide the missing proofs, the paper could be acceptable as a methodological note. The current absence of the B parameter value alone makes the main numerical claim unverifiable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Eric, quick take on Okyere et al. The paper does exactly what it says: takes the SIR and SEIR models from Rachah and Torres (2015, 2016), derives adjoint equations and control characterizations via Pontryagin's maximum principle, adds one extra SIR strategy (treatment plus education), and runs forward-backward sweep numerics. The algebra is correct for the stated Hamiltonians; we checked the adjoint systems and the control formulas — they're standard but right. The paper is honest about being a re-study, and the numerical conclusion matches what the model structure forces.\n\nWhat's new is the added SIR strategy and the explicit indirect-method treatment of the SEIR strategies, which is a useful check on the earlier direct-method results. For a textbook exercise, it's well organized.\n\nSoft spots: the omitted proofs are fine — Theorem 1 is proven and the others are truly similar. More annoying are the typos: system (5) has di/dt = ν s(t) s(t) instead of ν s(t) i(t), and equation (34) has a stray bracket. These are minor and recoverable. The bigger issue is the numerics for SIR Strategy 1: the objective (4) has weight B, but Section 2.3 never gives B a value; it lists D=1, C1=1, C2=5, C3=5, and D is the SEIR vaccination weight, not the SIR weight. So Figure 2's Strategy 1 trajectory and Figure 4's control are computed with an unreported parameter. No code or convergence criteria are supplied, so the central numerical claim is not independently reproducible. A single sentence with B=1 would fix this, but as written the evidence base is under-specified.\n\nI'd also flag the structural circularity: because vaccination/education move susceptibles to recovered and treatment moves infected to recovered, infection inevitably drops when those controls act. The conclusion that controls reduce transmission is not a finding; it's a property of the model. The authors don't overclaim — they say 'numerical illustrations' — but a reader shouldn't mistake this for an empirical result.\n\nWho's it for: someone teaching optimal control in epidemic modeling, or a researcher wanting a fully worked indirect-method counterpart to Rachah–Torres. It doesn't belong in a top journal, but it's a legitimate, complete manuscript. With the B issue fixed and the typos cleaned up, it's acceptable for a mid-tier applied math or modeling journal. I'd send it to review rather than desk reject; the referee can verify the derivations in an hour.","headline":"A competent but routine re-derivation of Rachah–Torres SIR/SEIR optimal control models via Pontryagin, with one new SIR strategy; the math checks out, but the main numerical claim is under-specified (missing B) and the conclusion is built into the model.","tokens_in":13639,"tokens_out":2399,"would_cite":false,"duration_ms":21079,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92D30","49K15","49J15"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that time-dependent vaccination, treatment, and educational campaigns reduce Ebola transmission in SIR and SEIR models, and it derives the optimal control schedules that do so.","keywords":["Ebola virus disease","SIR model","SEIR model","optimal control","maximum principle","vaccination","treatment","educational campaign"],"falsifier":"Run the uncontrolled SIR system with the paper's values ($\\nu=0.2$, $\\delta=0.1$) and compare its predicted epidemic curve with confirmed Ebola case counts from a real outbreak; if the model cannot reproduce the observed timing, peak, or final size, then the control comparisons built on the same equations are not quantitatively reliable for Ebola.","tokens_in":12664,"feed_emoji":"💉","tokens_out":10345,"duration_ms":91154,"temperature":0.7,"pith_summary":"The paper studies Ebola transmission through SIR and SEIR compartmental models in which vaccination, treatment, and educational campaigns act as time-dependent controls. Using the maximum principle, it derives adjoint equations and optimality systems for five control strategies, including one new SIR strategy that combines treatment with education. Numerical solutions of the optimal control problems show that each controlled strategy reduces the proportions of susceptible, exposed, and infected individuals and raises the recovered proportion compared with no control. The conclusion the authors draw is that effective education, vaccination, and treatment can help reduce Ebola transmission. The added SIR strategy is the paper's own extension to the earlier SIR and SEIR formulations.","feed_headline":"Vaccination, treatment, education curb modeled Ebola spread","feed_subtitle":"Optimal time-dependent controls lower infected and susceptible proportions in SIR and SEIR Ebola models.","key_machinery":"The load-bearing machinery is the controlled compartmental SIR and SEIR systems together with the maximum principle of optimal control. Each model is scaled by total population so states are proportions, and the controls enter as transfer rates: vaccination and education move susceptibles into the recovered class, and treatment moves infected individuals into the recovered class. For each strategy the paper constructs a Hamiltonian, derives the adjoint differential equations with zero terminal conditions, and obtains the optimal controls as clamped expressions of the adjoint differences, e.g., $\\eta^*(t)=\\min\\{1,\\max\\{0, s^*(t)(\\phi_1-\\phi_4)/D\\}\\}$ for the SEIR vaccination strategy. The forward-backward sweep scheme with fourth-order Runge-Kutta then solves the state and adjoint systems alternately until convergence, and it is this combined derivation-and-sweep procedure that produces the numerical comparisons.","core_discovery":"The central claim is that optimal time-dependent controls, applied to standard SIR and SEIR Ebola models, produce a clear reduction in disease burden relative to the uncontrolled systems. The paper establishes this by writing each model as a controlled dynamical system, minimizing an objective functional that balances infected counts against quadratic costs of vaccination, treatment, and education, and applying necessary optimality conditions to obtain adjoint equations and explicit formulas for the optimal controls. It adds one strategy to the SIR model on top of the earlier SIR and SEIR formulations, namely treatment combined with education, and it solves the resulting optimality systems numerically with a forward-backward sweep using fourth-order Runge-Kutta. The reported simulations show lower susceptible and infected proportions and higher recovered proportions under every controlled strategy, leading to the paper's stated conclusion that educational campaigns, vaccination, and treatment can reduce Ebola transmission.","pith_inferences":["The same machinery could be applied to Ebola models with quarantine, hospitalization, safe burial, or post-death transmission; those compartments would change the adjoint equations and likely change the optimal balance between education and treatment.","A natural testable extension is to fit the parameters to a real outbreak curve and compare the predicted optimal schedules with the interventions actually deployed, which would indicate whether the modeled savings are realistic.","The instant-transfer assumption for vaccination and education is an idealization; adding a delay between vaccination and protection, or between education and behavior change, would likely soften the reported reductions.","One could rank strategies by computing the total cost needed to achieve a fixed reduction in the final epidemic size, rather than only comparing trajectories."],"forward_implications":["The modeled epidemics peak with fewer infected individuals when the optimal schedules are applied, so the same conclusion holds for any Ebola-like disease with a similar transmission structure.","The additional SIR strategy of combining treatment with education expands the SIR intervention options from vaccination-only to also include a treatment-plus-education design.","The numerical agreement with direct-method results in prior work suggests the indirect optimal-control route is interchangeable with direct solvers for these models.","Because the objective functionals carry cost weights, the framework also yields information about intervention cost: cheaper controls are used at higher intensity, all else equal."],"supporting_citations":[{"why":"Supplies the SIR model, parameter values $\\nu=0.2$ and $\\delta=0.1$, and the vaccination-only SIR control formulation that this paper re-solves with indirect methods.","marker":"(Rachah and Torres, 2015)"},{"why":"Supplies the SEIR model, parameter values $\\nu=0.2$, $\\delta=0.1$, $\\rho=0.1887$, and the three SEIR strategies that this paper treats with the maximum principle.","marker":"(Rachah and Torres, 2016)"},{"why":"Provides the maximum principle used to derive every adjoint system and optimal control characterization.","marker":"(Pontryagin et al., 1962)"},{"why":"Describes the forward-backward sweep with fourth-order Runge-Kutta that generates the numerical solutions.","marker":"(Lenhart and Workman, 2007)"},{"why":"Grounds the compartmental SIR and SEIR modeling and the constant-population, no-vital-dynamics assumption.","marker":"(Hethcote, 2000)"}],"fun_headline_variants":["Ebola models show optimal controls lower infection","Time-dependent controls cut Ebola in SIR/SEIR models","Optimal control curbs Ebola: vaccination, treatment, education","Ebola spread reduced by optimal vaccination and treatment","Modeled Ebola tamed by dynamic control strategies"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the population is closed and constant in size and that vaccination, education, and treatment act by transferring people directly into the recovered class; if births, deaths, waning immunity, or delayed protection are substantial, the numerical trajectories no longer describe an Ebola outbreak.","fun_headline_variants_meta":{"raw":{"variants":["Ebola models show optimal controls lower infection","Time-dependent controls cut Ebola in SIR/SEIR models","Optimal control curbs Ebola: vaccination, treatment, education","Ebola spread reduced by optimal vaccination and treatment","Modeled Ebola tamed by dynamic control strategies"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000139,"raw_usage":{"total_tokens":1106,"prompt_tokens":841,"completion_tokens":265,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":457,"completion_tokens_details":{"reasoning_tokens":189}},"tokens_in":457,"tokens_out":265,"duration_ms":3074,"temperature":1.0,"reasoning_tokens":189,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:52:35.986828+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the uncontrolled SIR system with the paper's values ($\\nu=0.2$, $\\delta=0.1$) and compare its predicted epidemic curve with confirmed Ebola case counts from a real outbreak; if the model cannot reproduce the observed timing, peak, or final size, then the control comparisons built on the same equations are not quantitatively reliable for Ebola.","supporting_citations":[],"review_version":1}