{"id":"76a8fdd3-11b0-492e-bfbb-4f66fecd23dc","arxiv_id":"2507.01310","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A stochastic SIS model with immigration shows noise-induced bimodality and a mean-field Ising critical point, while vaccine hesitancy and vaccination data are described by a Beta distribution.","lead":"This paper analyzes a stochastic SIS epidemic model with immigration and random noise, finding that it undergoes noise-induced transitions and a phase transition of the mean-field Ising type. It also builds a vaccine-hesitancy model based on Kirman's ant model and reports that a Beta distribution fits COVID-19 vaccine willingness and vaccination data.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The Ising-universality claim rests entirely on the cubic normal form of the extrema equation; no steady-state moment or response function is computed, so the assignment to a universality class of the stochastic process is not tested.","rationale":"The reader's conditional verdict already captures the main issue: the Fokker-Planck algebra is sound, but the universality-class claim is not independently verified. My reading adds precision: Eq. (20) makes the cubic normal form match the Ising equation of state by definition, so Fig. 7 is circular as a numerical test. This does not force rejection; the derivation of the steady-state PDF, the cusp coordinates, and the vaccination threshold Rc = 1 + sigma^2/(2 gamma) appear internally consistent. However, the central claim that the transition belongs to the mean-field Ising universality class is load-bearing and would require a direct computation of moment scaling or of a genuine response function. The required revision is to compute <y>, the variance, and d<y>/dh from Eq. (10) near the cusp and compare their exponents with Eqs. (22)-(24). Until such a calculation is supplied, the classification remains a normal-form analogy rather than a property of the stochastic process. Therefore the reader's CONDITIONAL verdict is appropriate and no change is needed.","tokens_in":18466,"tokens_out":6609,"duration_ms":82058,"concrete_test":"Compute numerically from Eq. (10) the steady-state moments <y>=integral y ps(y) dy and <y^2> near the cusp (R0,V)=(1.73854,0.516404) with R1=0.01, along the tangential line V=0.594065 R0 - 0.516402 and along the coexistence line. Fit |<y>-yc| and the variance to |theta| as theta approaches 0, using the theta defined in Eq. (20). If the mean exponent is not 1/2 or the variance does not diverge as |theta|^{-1}, the claim that the stochastic process belongs to the mean-field Ising universality class is not supported. As an additional check, add a small field-like term -h y to f(y) in Eq. (7), recompute ps(y), and measure d<y>/dh at h=0; this would test gamma_c=1.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central classification in Section III uses Eq. (20) to map the Taylor expansion of F(y)=0 onto the mean-field Ising equation of state. By construction, h and theta are chosen as rescaled values of F(yc) and F'(yc), so the cubic equation (19) is algebraically identical to Eq. (18); Fig. 7, which evaluates Eq. (20), therefore cannot provide independent evidence for beta_c=1/2. The order parameter m=(ym-yc)/yc is the displacement of the most probable value, not a thermal average. The steady-state Fokker-Planck PDF Eq. (10) is explicit, but the paper never computes <y>, the variance, or any susceptibility of the stochastic process near the cusp. At a cusp of a PDF, the maxima can undergo a pitchfork while the mean follows the relative weights of the two peaks, so moments can scale with different exponents. gamma_c=1 is especially unsupported: chi=(dm/dh) is defined through the formal mapping, not through an actual response of the process to a perturbing field. Thus the universality-class statement is currently a statement about the Landau normal form, not about the stochastic SIS model.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a stochastic SIS epidemic model with immigration and multiplicative noise. It derives the stationary Fokker-Planck probability density, the extrema equation F(y)=0, and the R0-V bifurcation diagram showing a cusp point. By Taylor-expanding the extrema equation around the inflection point and comparing with the mean-field Ising equation of state, it claims the cusp transition belongs to the mean-field Ising universality class with exponents beta_c=1/2, gamma_c=1, delta=3. The paper also analyzes vaccination: it finds a stochastic disease-elimination threshold Rc=1+sigma^2/(2gamma), proposes a Kirman-type vaccine-hesitancy model leading to a Beta stationary distribution, fits the Beta distribution to COVID-19 vaccine-willingness and vaccination-coverage data, and derives a distribution for the effective reproduction number under a Beta-distributed vaccinated fraction.","tokens_in":18600,"tokens_out":9975,"duration_ms":109119,"significance":"The analytic steady-state calculations are clean and the explicit connection between the SIS-with-immigration bifurcation diagram and the Horsthemke-Lefever model is a useful contribution. The threshold shift Rc = 1 + sigma^2/(2gamma) is a concrete, potentially testable prediction that goes beyond the deterministic result. However, the universal-features claim is currently a statement about the Landau normal form of the extrema equation rather than about observables of the stochastic process, and the empirical fits in Sections V-VI rely on statistically invalid KS procedures. With appropriate reframing and, ideally, additional moment-based checks, the result would be of genuine interest to the stochastic-bifurcation and epidemic-modeling communities.","major_comments":[{"comment":"The assignment of the mean-field Ising universality class is not tested for the stochastic process. The variable m=(ym-yc)/yc is the displacement of the most probable value from the inflection point of the PDF, not a thermal average; h and theta are defined directly from F(yc) and F'(yc) in Eq. (20), so Eq. (19) is just the Taylor expansion of the same equation F(y)=0, and the exponents beta_c=1/2, gamma_c=1, delta=3 follow algebraically from the cubic normal form. Figure 7 therefore evaluates Eq. (20) and cannot provide independent evidence for beta_c=1/2. The authors should either explicitly state that the claim is about the mean-field normal form of the extrema equation, or compute actual stationary moments (mean <y>, variance, and a response/susceptibility of the Fokker-Planck steady state) near the cusp to demonstrate that the stochastic process itself exhibits these exponents.","section":"Section III, Eqs. (18)-(24) and Fig. 7"},{"comment":"The susceptibility exponent gamma_c=1 is particularly unsupported. The quantity chi=(dm/dh) is a derivative in the formal mapping of Eq. (20); no physical field conjugate to the infected fraction or to the mode location is identified, and no fluctuation-dissipation relation for the Fokker-Planck model is provided. As written, gamma_c is a property of the expanded cubic equation, not a measurable response of the SIS model. Please justify this exponent as a property of the epidemic model or remove it from the universality-class claim.","section":"Section III, Eq. (23)"},{"comment":"The disease-free probability pst(r <= rth) is computed with rth=1/R0, i.e., the deterministic transcritical threshold R=1, whereas Section IV shows that in the stochastic model the threshold is Rc=1+sigma^2/(2gamma) (Eq. (26)). If the disease-free condition is to be evaluated in the stochastic framework of the paper, the threshold in Eq. (37) and in Fig. 12 should be the stochastic threshold; otherwise the vaccine-hesitancy conclusions are tied to a deterministic model and this needs to be stated explicitly.","section":"Section VI, Eqs. (35)-(37)"},{"comment":"The KS-test evidence for the Beta distribution is not statistically valid as presented. The distribution parameters are estimated by MLE from the same small samples (19-23 countries for the hesitancy data), and the standard one-sample KS null distribution used by 'kstest' does not apply when parameters are estimated; a Lilliefors-type correction or a simulation-based calibration is required. In addition, the histograms use arbitrary bin widths and the sample sizes are not reported. Consequently, the p-values in Tables II and III cannot support the claim that the Beta distribution (or any distribution) fits the data well, and the empirical motivation for Section VI rests on weak grounds.","section":"Section V, Tables I-III"},{"comment":"The statement that a Beta-distributed q makes the bifurcation diagram probabilistic is heuristic. If q is random while the infection dynamics are stochastic, one must formulate a joint Fokker-Planck equation for (y,q) or state the simplifying assumptions under which the Beta distribution of q can be superimposed on the bifurcation diagram. As written, Eq. (35) is only a change of variables for fixed R0 and is not a derivation from the epidemic dynamics; the status of this paragraph should be clarified.","section":"Section VI, final paragraph"}],"minor_comments":[{"comment":"Equation (19) contains a typographical error: the third-order term should be F'''(yc)(y-yc)^3/3!, not 'F'''(y-yc)(y-yc)^3/3!'.","section":"Section III, Eq. (19)"},{"comment":"The distribution pst(R) in Eq. (35) is the distribution of the effective reproduction number R=R0(1-q), not of the basic reproduction number R0. The abstract and surrounding text should use 'effective reproduction number' consistently.","section":"Section VI, Eq. (35)"},{"comment":"The caption should state explicitly that the plot is computed directly from the mapping in Eq. (20), i.e., it is an algebraic consistency check of the normal form rather than a numerical simulation or an independent measurement.","section":"Figure 7 caption"},{"comment":"The text says the best fits in 2020 and 2021 are Normal and Weibull, respectively, while the concluding sentence emphasizes Beta; the narrative would be clearer if it explicitly stated that the Beta distribution is competitive but not always the best by the KS statistic.","section":"Section V, Table II"},{"comment":"The notation R1 is used both for the immigration ratio eta/gamma in Section III and for a generic variable in the vaccine-hesitancy model in Section V; different symbols for these unrelated quantities would avoid confusion.","section":"Section II and III"}],"recommendation":"major_revision","confidential_remarks":"The paper's central Fokker-Planck derivation is sound and the vaccination threshold Rc is an interesting result, but the title-level claim of Ising universality is currently a normal-form statement rather than a demonstrated property of the stochastic process. I would not reject the manuscript, but the revision must address the normal-form versus process-level distinction and the statistical validity of the KS tests. If the authors can either narrow the universality claim or provide moment-based evidence, the paper would be publishable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this paper as a clean mean-field analysis of a stochastic SIS model with immigration and multiplicative noise. The Fokker-Planck steady state, extrema equation, saddle-node boundaries, and cusp coordinates are all explicit and checkable. The connection to the Horsthemke-Lefever model is real, and the vaccination threshold Rc = 1 + sigma^2/(2gamma) is a useful extension of the known stochastic SIS result to the vaccinated case. If you work in noise-induced transitions, this is worth a look.\n\nThe new mathematical content is the specific SIS-with-immigration realization of the known HL-type bifurcation diagram, with the cusp point computed in closed form. That part is solid. The vaccine-hesitancy section applies Kirman's ant model and fits the resulting Beta distribution to survey data; that's not a new model, but the data application is reasonable in spirit.\n\nThe soft spots are real but not fatal. The Ising-universality claim is the biggest one. The exponents beta=1/2, gamma=1, delta=3 are read off the cubic Taylor expansion of the extrema equation after mapping to the mean-field equation of state. That is a normal-form statement, not a statement about the stochastic process. The paper never computes the mean infected fraction, the variance, or a response function near the cusp for the eta≠0 case, and Figure 7 is self-referential because it evaluates the same mapping. Also, Figure 9 shows <y> for eta≠0 along a line that does not pass through the cusp, so it does not test the claimed scaling. I would rephrase to: 'the cusp is described by the mean-field Ising normal form' and leave 'universality class' out unless moment-based or simulation evidence is supplied. This is a fixable wording issue, not a collapsed derivation.\n\nThe statistical section is more problematic. The KS tests are applied to histograms with arbitrarily chosen bin sizes, on samples of 19-23 countries, using parameters estimated from the same data. The p-values are therefore not valid as reported. A Lilliefors-type correction or parametric bootstrap is needed. The Section VI 'distribution of R0' is just the change of variables from the fitted Beta distribution on q; calling it a 'derivation of the steady-state distribution of R0' oversells it.\n\nOverall: the core mean-field math is honest and checkable, and the paper cites the relevant prior work, including the Gray et al. proof of the no-vaccination version of Eq. (26). With rewording of the universality claim and a proper goodness-of-fit procedure, this is suitable for publication. It deserves peer review, not desk rejection.\n\nReferee recommendation: engage with it; ask for the rewording and corrected statistics.","headline":"Solid mean-field analysis of stochastic SIS with immigration; the Ising-universality label needs support or rewording, and the vaccine-data fitting needs statistical fixes, but the core math is checkable and worth a serious referee.","tokens_in":19275,"tokens_out":4913,"would_cite":true,"duration_ms":52179,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92D30","60H10","82B27"],"pacs":[],"model":"deepseek-v4-flash","headline":"Adding immigration to the SIS epidemic model and driving it with multiplicative noise produces a cusp point in the bifurcation diagram where the transition between unimodal and bimodal stationary distributions belongs to the mean-field…","keywords":["stochastic SIS model","noise-induced transition","Fokker-Planck equation","mean-field Ising universality","cusp bifurcation","vaccination threshold","Beta distribution","vaccine hesitancy"],"falsifier":"Compute from the Fokker-Planck equation the stationary mean infected fraction $\\langle y\\rangle$, its variance, and a response function as the cusp is approached along the tangential line with the reduced field $h=0$, and check whether $\\langle y\\rangle-y_c$ scales as $(-\\theta)^{1/2}$ and the response diverges as $\\theta^{-1}$ with the predicted amplitudes; if the moments do not show these power laws, the universality claim for the process is falsified even if the normal-form mapping holds.","tokens_in":18135,"feed_emoji":"🦠","tokens_out":10391,"duration_ms":102703,"temperature":0.7,"pith_summary":"The paper studies a stochastic susceptible-infected-susceptible (SIS) epidemic model with immigration, in which environmental fluctuations multiply the infection rate. It claims that the steady-state probability distribution of the infected fraction undergoes noise-induced transitions between unimodal and bimodal shapes, and that the boundaries of bimodality end in a cusp point at which the transition is critical. At that cusp, the paper maps the extremum equation of the stationary distribution onto the mean-field Ising equation of state and concludes that the transition has mean-field Ising exponents $\\beta_c=1/2$, $\\gamma_c=1$, $\\delta=3$. In the vaccination version without immigration, the analysis yields a disease-elimination threshold $R_c=1+\\sigma^2/(2\\gamma)$, so stochasticity raises the threshold and lowers the vaccination coverage needed for eradication. The paper also derives a Beta distribution for vaccine willingness from a herding ant-colony model and fits it to COVID-19 vaccine acceptance and global vaccination data.","feed_headline":"Noise shifts epidemic threshold and eases vaccination burden","feed_subtitle":"A stochastic SIS model also shows a mean-field Ising cusp transition in its bifurcation diagram.","key_machinery":"The load-bearing object is the steady-state probability density $p_s(y)$ of the infected fraction, obtained in closed form from the Fokker-Planck equation, and its extremum equation $F(y_m)=0$. The extrema $y_m$ serve as the effective state variable; the number of maxima distinguishes unimodal from bimodal regimes. At the cusp, the Taylor expansion of $F$ around the inflection point $y_c=1/2-1/(6V)$ is cubic with no quadratic term, so it can be rescaled into the magnetic equation of state $h-\\theta m-m^3/3=0$, making the cusp transition mean-field Ising-like. For vaccination, the threshold follows from the upper bifurcation line $V_R=2(1-1/R)$ combined with the line $R V_R = \\sigma^2/\\gamma$, giving $R_c=1+\\sigma^2/(2\\gamma)$. For vaccine hesitancy, the machinery is a two-state herding model whose continuum Fokker-Planck equation has as steady state the Beta distribution.","core_discovery":"The central claim is that the stochastic SIS model with immigration, under multiplicative noise, has a critical-point transition at a cusp point of its $V$-versus-$R_0$ bifurcation diagram, and that this transition belongs to the mean-field Ising universality class. The order parameter is the location $y_m$ of the maximum of the steady-state probability density; expanding the extremum condition $F(y_m)=0$ around the inflection point $y_c$ removes the quadratic term and gives the Landau normal form $h-\\theta m-m^3/3=0$, yielding exponents $\\beta_c=1/2$, $\\gamma_c=1$, and $\\delta=3$. A secondary claim is that in the SIS model with vaccination and no immigration, the disease-free absorbing phase is reached when the effective reproduction number $R$ falls below $R_c=1+\\sigma^2/(2\\gamma)$, which is larger than the deterministic threshold $1$ and reduces the critical vaccination coverage by $\\sigma^2/(2R_0\\gamma)$. The paper further claims that the vaccine-hesitancy dynamics, modeled after an ant-colony herding process, has a Beta-distributed steady state for the vaccine-willing fraction, and that the same Beta distribution describes COVID-19 vaccine willingness and global vaccination data.","pith_inferences":["A decisive test of the Ising-universality claim, going beyond the paper, is to compute the stationary mean and variance of $y$ directly from the Fokker-Planck equation near the cusp; if the moment-based order parameter does not scale as $(-\\theta)^{1/2}$, the universality class applies to the Landau normal form rather than to the stochastic process's observables.","The threshold $R_c=1+\\sigma^2/(2\\gamma)$ suggests that environmental noise can partially substitute for vaccination, which an agent-based simulation with fluctuating transmission rates could test at the individual level.","The ant-colony herding model gives a mechanistic origin for the empirically popular Beta prior in Bayesian vaccine studies; coupling the herding dynamics with the epidemic itself, so that the vaccinated fraction co-evolves with infections, is a natural extension.","Because the cusp normal form is identical to that of the classic population-genetics noise model, the same argument should produce Ising-type criticality in other population and opinion dynamics with multiplicative noise and a linear immigration-like term."],"forward_implications":["In the stochastic SIS model with immigration, the cusp transition has exponents $\\beta_c=1/2$, $\\gamma_c=1$, and $\\delta=3$, so the noise intensity plays the role of temperature and the most probable infected fraction plays the role of magnetization.","In the vaccination version without immigration, total disease eradication occurs when the effective reproduction number $R=R_0(1-q)$ falls below $R_c=1+\\sigma^2/(2\\gamma)$; the required vaccination coverage is therefore $q_c=1-1/R_0-\\sigma^2/(2R_0\\gamma)$, which is lower than the deterministic value.","Even when $R>R_c$, all $R$-$V_R$ trajectories pass through regions of the bifurcation diagram where the steady-state PDF has a maximum at $y=0$, so there is a finite probability that the infection is eradicated.","The vaccine-hesitancy model gives a Beta distribution for the fraction of vaccine-willing individuals, and the Beta distribution provides the best fit among the tested distributions for the 2022 COVID-19 vaccine and booster data and for global vaccination coverage data.","When the vaccinated fraction is Beta-distributed, the effective reproduction number has a scaled Beta distribution, and the probability of a disease-free state is given by an incomplete Beta function ratio that increases as the vaccine-acceptance parameter grows relative to the hesitancy parameter."],"supporting_citations":[{"why":"Supplies the deterministic SIS-with-immigration model and the vaccination parameterization used throughout.","marker":"[6]"},{"why":"Provides the canonical theory of noise-induced transitions and the mean-field Ising equation of state used in the mapping.","marker":"[7]"},{"why":"Supplies the magnetic equation of state and the scaling relations quoted for the critical exponents.","marker":"[9]"},{"why":"Provides the stochastic SIS model without immigration and its integrability condition, used as a special case in the bifurcation analysis.","marker":"[10]"},{"why":"Provides the procedure of expanding around the inflection point to cast the extremum equation in Ising normal form.","marker":"[13]"},{"why":"Supplies the herding ant-colony model on which the vaccine-hesitancy model is based.","marker":"[19]"},{"why":"Provides the COVID-19 vaccine willingness survey data used for the Beta-distribution fits.","marker":"[22]"},{"why":"States and proves the stochastic SIS theorem giving the shifted disease-elimination threshold.","marker":"[23]"},{"why":"Provides the evidence and beta-regression analysis connecting vaccine hesitancy to vaccine uptake.","marker":"[27]"},{"why":"Provides the global COVID-19 vaccination coverage data used for the Beta-distribution fit.","marker":"[28]"}],"fun_headline_variants":["Epidemic model's cusp transition joins Ising universality class","Noise lowers epidemic threshold, reducing vaccine coverage needed","Beta distribution fits vaccine hesitancy and vaccination data","Mean-field Ising transition found in epidemic model","Stochastic SIS model reveals universal critical behavior"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claim that the cusp transition is mean-field Ising rests on the assumption that the location of the maximum of the steady-state PDF, together with its cubic Taylor expansion near the inflection point, completely determines the critical behavior, while the paper never computes the actual moment-based order parameter or susceptibility of the stochastic process.","fun_headline_variants_meta":{"raw":{"variants":["Epidemic model's cusp transition joins Ising universality class","Noise lowers epidemic threshold, reducing vaccine coverage needed","Beta distribution fits vaccine hesitancy and vaccination data","Mean-field Ising transition found in epidemic model","Stochastic SIS model reveals universal critical behavior"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000499,"raw_usage":{"total_tokens":2491,"prompt_tokens":1041,"completion_tokens":1450,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":657,"completion_tokens_details":{"reasoning_tokens":1372}},"tokens_in":657,"tokens_out":1450,"duration_ms":14138,"temperature":1.0,"reasoning_tokens":1372,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T20:56:24.666565+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute from the Fokker-Planck equation the stationary mean infected fraction $\\langle y\\rangle$, its variance, and a response function as the cusp is approached along the tangential line with the reduced field $h=0$, and check whether $\\langle y\\rangle-y_c$ scales as $(-\\theta)^{1/2}$ and the response diverges as $\\theta^{-1}$ with the predicted amplitudes; if the moments do not show these power laws, the universality claim for the process is falsified even if the normal-form mapping holds.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the deterministic SIS-with-immigration model and the vaccination parameterization used throughout."},{"cited_title":"Horsthemke and R","cited_arxiv_id":null,"evidence_quote":"Provides the canonical theory of noise-induced transitions and the mean-field Ising equation of state used in the mapping."},{"cited_title":"Vespignani, Modelling dynamical processes in complex socio-technical systems, Nature Physics 8, 32 (2012)","cited_arxiv_id":null,"evidence_quote":"Supplies the magnetic equation of state and the scaling relations quoted for the critical exponents."},{"cited_title":"In the case of the SIS model with no immigration, Eq","cited_arxiv_id":null,"evidence_quote":"Provides the stochastic SIS model without immigration and its integrability condition, used as a special case in the bifurcation analysis."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the procedure of expanding around the inflection point to cast the extremum equation in Ising normal form."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the herding ant-colony model on which the vaccine-hesitancy model is based."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the COVID-19 vaccine willingness survey data used for the Beta-distribution fits."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"States and proves the stochastic SIS theorem giving the shifted disease-elimination threshold."},{"cited_title":"Franceschi, L","cited_arxiv_id":null,"evidence_quote":"Provides the evidence and beta-regression analysis connecting vaccine hesitancy to vaccine uptake."},{"cited_title":"Kirman, Ants, Rationality, and Recruitment, The Quar- terly Journal of Economics 108, 137 (1993)","cited_arxiv_id":null,"evidence_quote":"Provides the global COVID-19 vaccination coverage data used for the Beta-distribution fit."}],"review_version":1}