Pith. sign in

REVIEW 5 major objections 4 minor 44 references

From Risk Perception to Behavior Large Language Models-Based Simulation of Pandemic Prevention Behaviors

T0 review · 5 major / 4 minor · reviewed 2026-08-03 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that a large language model, prompted from a first-person perspective, can simulate how populations adopt and drop pandemic prevention behaviors—matching observed survey distributions with 72.7% zero-shot accuracy, 81.8% w

desk verdict A plausible LLM-based simulation framework for pandemic behaviors, but the validation is too lenient to back the headline accuracy numbers. read the letter →

arxiv 2601.03552 v2 pith:PXSLQ5DR submitted 2026-01-07 cs.SI

Lujia Bo , Mingxuan Chen , Youduo Chen , Xiaofan Gui , Jiang Bian , Chunyan Wang , Yi Liu This is my paper · ORCID
classification cs.SI
keywords largelanguagemodelspreventionbehaviorsimulationriskperceptionfew-shotlearningcross-contexttransferCOVID-19Kolmogorov-Smirnovtestpolicy
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that LLM-based simulation can stand in for traditional survey-driven behavioral models when data are scarce, by having the model reason through a resident's risk perception and habit formation. It builds two coupled prompts: one that maps demographic and epidemic-policy context to probabilities for eleven specific behaviors, and one that updates perceived risk over time and feeds the updated risk back into the static module. Validation against two rounds of Beijing surveys shows match rates of 72.7%, 81.8%, and 77.8% under zero-shot, few-shot, and cross-context transfer, with the Kolmogorov-Smirnov test at a permissive threshold as the acceptance criterion. If these numbers hold, epidemic response teams could use the framework to stress-test behavioral responses long before empirical data arrive, and to see which behaviors persist or collapse when policies relax.

What carries the argument

The load-bearing device is a five-part 'first-person perspective' prompt: basic demographic identity, pandemic context (R0, CFR, severity), community control measures, an environmental risk perception score, and a task asking for a 0-1 execution probability for each of eleven behaviors plus a reasoning rationale. A second prompt takes the T1 risk perception and the T1-to-T2 changes in pandemic and policy context, and outputs an updated risk score; that score is then fed into the static module to produce behavioral evolution. Ten simulation rounds are averaged per condition to dampen stochasticity, and the outputs are discretized into Likert scales for comparison with survey data. The Kolmogo

What would settle it

Re-run the same simulated-versus-observed comparisons using alpha=0.05 or an effect-size metric like Cohen's d or the Cramer-von Mises distance, or compare the model's output to a baseline that always predicts the training marginal distribution. If most behaviors then fail to match, or the accuracy numbers drop toward chance, the paper's claim of predictive validity is not supported.

Watch

Extended reading notes

Core claim

The central claim is that a structured, first-person prompt can make a large language model reproduce the distribution of eleven prevention behaviors measured on five-point Likert scales, in both a static snapshot and a dynamic update where perceived risk is revised and behavior re-simulated. The paper reports that under zero-shot conditions eight of eleven behaviors match the observed Beijing R1 distributions; with one-third of the data as few-shot references nine of eleven match; and in cross-context transfer from R1 to the later R2 wave seven behaviors match. The same machinery, applied to China's December 2022 policy relaxation, yields broad behavioral loosening but a sustained increase

Load-bearing premise

The entire validity chain rests on passing a Kolmogorov-Smirnov test at the unusually lenient threshold of p>0.001, which counts even fairly rough distribution matches as successes.

Editorial extensions

If this is right

  • With only one-third of one survey wave as reference examples, the framework matches observed behavior distributions for nine of eleven behaviors, suggesting minimal local data may suffice for calibration.
  • Cross-context transfer from an earlier to a later epidemic wave maintains a 77.8% match rate, implying historical data can be repurposed for novel policy contexts.
  • The simulated December 2022 relaxation shows most behaviors declining while drain-seal disinfection intensifies, a counter-trend the paper links to low-cost habits and sewage-related risk perception.
  • Across 120 epidemic scenarios, behavior intensity under no community control can exceed intensity under strong control, indicating that control stringency and perceived risk do not move in lockstep.
  • The framework's decision rationales provide interpretable, policy-diagnostic output, not just distribution match.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial extension: the transfer numbers may be optimistic because the hierarchical protocol only lets behaviors that pass few-shot validation advance to transfer testing, which filters out the hardest-to-predict behaviors before the 77.8% figure is computed.
  • If the match rates are real, the same prompting recipe could pre-screen behavioral responses in cities lacking survey data, but only where the model's priors align with the local culture; transfer to non-Chinese contexts is untested here.
  • The drain-disinfection finding suggests an environmental externality that the paper quantifies as hundreds of thousands of tons of disinfectant per year for Beijing; public-health guidance may need to consider wastewater by-products when advocating low-cost disinfection.
  • A stricter significance threshold (p>0.05) or a distribution-distance effect size would give a more decisive test of whether the model is truly reproducing behavior or merely approximating a broad shape.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 4 minor

Summary. The paper proposes an LLM-based framework for simulating pandemic prevention behaviors, combining a static module that maps demographic, epidemic-policy, and risk-perception inputs to behavior intensities and a dynamic module that updates perceived risk over time. The framework is evaluated against two Beijing survey waves (R1, R2) under zero-shot, few-shot, and cross-context transfer strategies, with a Kolmogorov-Smirnov test at p>0.001 used as the validity criterion. Headline predictive-accuracy figures are 72.7% (zero-shot), 81.8% (few-shot), and 77.8% (transfer). The paper also presents simulations of behavior change during China's December 2022 policy relaxation and across 120 epidemic-condition scenarios.

Significance. If the validation were sound, the framework would be a valuable low-data tool for forecasting prevention behaviors in novel outbreak contexts, and the use of real survey data, an explicit progressive validation protocol, and interpretable LLM rationales are genuine strengths. However, the current statistical evidence is weak: the validation threshold is unusually lenient, no baseline model is reported, and the transfer result is computed on a pre-selected subset. The central quantitative claims are therefore not established by the presented analyses.

major comments (5)
  1. [Performance Evaluation] The central validation uses KS p>0.001 as a 'validity criterion.' This is an extremely lenient bar: non-rejection at α=0.001 provides weak evidence of distributional equivalence, especially with n=120 and 5-point Likert data, where the KS test is conservative in the presence of ties and has low power. The paper does not report KS statistics, effect sizes, power analysis, or equivalence bounds. The headline 'predictive accuracy' is simply the fraction of behaviors passing this threshold, so the percentages are not interpretable without a null-model comparison. Please report D values, use a pre-specified conventional threshold or equivalence margin, and demonstrate what a trivial model would achieve.
  2. [Model Evaluation Strategies] No baseline model is included. The few-shot condition provides one-third of R1 as reference examples, and the static module receives survey-measured risk perception as a direct input, which is known to correlate with protective behavior. A simple empirical distribution from the reference examples, or a logistic regression using risk perception alone, could plausibly pass many of the same KS tests. The Discussion's claim that GPT-4o achieves comparable results with '98% less data than gradient boosting models' is not supported by any comparative experiment in this manuscript.
  3. [Static Prevention Behavior Simulation] The static module directly conditions on the respondent's measured environmental risk perception via the 'Environmental Risk Perception' prompt component (see also Variable Construction and Prompt Input Processing). Since risk perception is a well-established predictor of prevention behavior, this setup may only demonstrate that the LLM can translate a given risk-perception score into behavior, rather than simulating behavior from external context from first principles. An ablation without the risk-perception input and a comparison against a risk-perception-only statistical model are needed to establish the framework's added value.
  4. [Model Evaluation Strategies] The transfer result of 77.8% is computed after a hierarchical filter: only behaviors that passed the few-shot strategy are advanced to Strategy 3. This makes the transfer percentage conditional on earlier success and not comparable with the zero-shot and few-shot percentages. The paper does not report how many behaviors were excluded at the filtering step or the transfer performance for the full set of behaviors. Without this information, the 77.8% figure is potentially inflated and should be re-reported with full-cohort results.
  5. [Data Sources and Preprocessing] The paper uses propensity score matching to create 120 'matched pairs' between R1 and R2, but matching does not create longitudinal data. The dynamic module is then validated on these pseudo-transitions (R1-R2) and claims to predict behavior evolution over time. Since the same individuals were not followed, the dynamic evaluation cannot distinguish true within-person change from cross-sectional cohort differences. At minimum, this structural limitation should be acknowledged explicitly and the dynamic validation rephrased as a cross-sectional comparison, not a prediction of individual-level temporal evolution.
minor comments (4)
  1. [Performance Evaluation] The text cites 'Chen et al., 2017' to justify the KS threshold, but this reference does not appear in the reference list. Please provide the source or remove the citation.
  2. [Results] The abbreviations i, s, r in Table 1 and Figures 2-3 are defined but hard to parse in the main text. Consider spelling out 'Isolation,' 'Self-Health Monitoring,' and 'Regular P&C' in figure captions or table notes for clarity.
  3. [Case Applications] The R3 policy-relaxation and 120-condition simulations are not empirically validated. They should be explicitly framed as hypothetical scenario analyses rather than predictive results, since no ground-truth data are available.
  4. [Implementation and Computational Environment] The prompt templates and code are not provided, which limits reproducibility for a prompt-engineering-driven framework. Including them in supplementary material would substantially strengthen the paper.

Circularity Check

0 steps flagged · score 0.0 of 10

No formal circularity; the reported weaknesses are statistical validation concerns rather than derivation-by-construction.

full rationale

The paper's claimed derivation chain is not circular in the sense defined here. The static module is a prompted LLM that takes demographic attributes, pandemic context, community control measures, and an environmental risk perception state as inputs and produces behavior probabilities; these outputs are not manipulated algebraically from the target survey distributions. The validation compares simulated 5-point Likert distributions to observed survey distributions via a KS test; this is a statistical decision rule, not an equation that forces the simulated output to equal the observed output. The few-shot condition uses one-third of R1 as in-context examples and tests on the remaining two-thirds; the transfer condition uses R1 data to predict R2, an independent survey round. The dynamic module uses observed T1 risk perception as an initial state and simulates T2 behavior; this is a conditional simulation, not a definitional identity. The hierarchical protocol, under which only behaviors that passed few-shot are advanced to transfer, inflates the reported transfer percentage but does not make any prediction equivalent to its input. Self-citations to Li et al. (2025) provide the empirical data and interpretive context (e.g., stronger sewage-related risk-perception links); they are not invoked as an unverified uniqueness theorem or ansatz that carries the derivation. The use of survey-measured risk perception as a static-module input is a potential label-leakage confound and weakens the 'prediction from external context' framing, but it is not an equation-level circularity. Likewise, the lenient p>0.001 KS threshold and absence of baselines are serious validity concerns; per the review rules, those belong to correctness risk rather than circularity. Therefore no circular step is identified.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central empirical claims depend on several hand-chosen thresholds and assumptions: the very permissive KS validity criterion, the selection of few-shot examples, discretization choices, and the hierarchical pre-filtering of behaviors. The theoretical background (risk perception as a driver, LLM-as-human proxies) is assumed from prior work rather than independently verified here. No new physical or existence claims are made.

free parameters (5)
  • KS validity threshold = 0.001
    Chosen by hand; p-values above this threshold count as validation success. This is a much lower bar than conventional alpha=0.05, making the test extremely permissive.
  • Few-shot reference examples = one-third of R1 (approx. 327 respondents)
    Randomly sampled one-third of R1 data as in-context examples; affects few-shot and transfer results.
  • Discretization thresholds = equidistant intervals for 5-point and 6-point Likert scales
    Continuous LLM outputs are binned using equidistant thresholds, a modeling choice that affects all downstream comparisons.
  • R3 scenario parameters = R0=10.0, CFR=0.05%
    Hand-set values for the December 2022 policy relaxation simulation; not derived from the model.
  • Hierarchical selection rule = advance to transfer only if passed few-shot
    Behaviors are pre-filtered before transfer evaluation, inflating reported transfer accuracy.
assumptions (5)
  • domain assumption Risk perception is a direct driver of prevention behavior
    Borrowed from Bauer (1960) and Slovic (1987); the model builds on this premise without testing it.
  • domain assumption LLMs can simulate human cognitive and behavioral processes accurately
    Relies on cited works (e.g., Binz & Schulz, Horton) for validity; not independently verified here.
  • ad hoc to paper KS test with p>0.001 is an appropriate validity criterion
    The threshold is set unusually low; standard practice uses alpha=0.05, making the criterion very permissive.
  • domain assumption PSM-matched samples are comparable
    Assumes propensity score matching on demographics makes R1 and R2 samples comparable for longitudinal inference.
  • ad hoc to paper Equal-width discretization preserves distributional information
    No justification for equidistant bins over alternative binning; affects all reported distributions.

how reviews work

0 comments
Cite this review

Pith. "Pith review of From Risk Perception to Behavior Large Language Models-Based Simulation of Pandemic Prevention Behaviors." pith.science (2026). https://pith.science/paper/PXSLQ5DR

@misc{pith2026260103552,
  author       = {Pith},
  title        = {Pith review of: From Risk Perception to Behavior Large Language Models-Based Simulation of Pandemic Prevention Behaviors},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PXSLQ5DR}},
  note         = {Machine review of arXiv:2601.03552}
}
read the original abstract

Individual prevention behaviors are a primary line of defense during the early stages of novel infectious disease outbreaks, yet their adoption is heterogeneous and difficult to forecast-especially when empirical data are scarce and epidemic-policy contexts evolve rapidly. To address this gap, we develop an LLM-based prevention-behavior simulation framework that couples (i) a static module for behavior-intensity prediction under a specified external context and (ii) a dynamic module that updates residents' perceived risk over time and propagates these updates into behavior evolution. The model is implemented via structured prompt engineering in a first-person perspective and is evaluated against two rounds of survey data from Beijing residents (R1: December 2020; R2: August 2021) under progressively realistic data-availability settings: zero-shot, few-shot, and cross-context transfer. Using Kolmogorov-Smirnov tests to compare simulated and observed behavior distributions (p > 0.001 as the validity criterion), the framework demonstrates robust performance and improves with limited reference examples; reported predictive accuracy increases from 72.7% (zero-shot) to 81.8% (few-shot), and remains high at 77.8% under transfer to novel contexts. We further apply the framework to simulate behavior changes during China's December 2022 policy relaxation and to stress-test behavioral responses across 120 systematically varied epidemic conditions (R0, CFR, and control-measure tiers). Results indicate broad behavioral loosening under relaxation but a distinctive counter-trend increase in drain-related disinfection, highlighting how low-cost, low-friction behaviors may persist or intensify even when external constraints recede-raising a potential environmental tradeoff.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 15 canonical work pages

  1. [1]

    K., Duek, E., Spiller,T.R., Perez,Y.,Wolf, R., Bhonker, S., Suslo, D., Burrer, A., & Schulz, E

    Ben-Zion, Z.,Witte, K., Jagadish,A. K., Duek, E., Spiller,T.R., Perez,Y.,Wolf, R., Bhonker, S., Suslo, D., Burrer, A., & Schulz, E. (2025). Assessing and alleviatingstateanxietyinlargelanguagemodels.npjDigitalMedicine,8,Article 132.https://doi.org/10.1038/s41746-025-01512-6

    work page doi:10.1038/s41746-025-01512-6 2025

  2. [2]

    Proceedings of the National Academy of Sciences, 120(6), e2218523120

    Binz,M.,&Schulz,E.(2023).UsingcognitivepsychologytounderstandGPT-3. Proceedings of the National Academy of Sciences, 120(6), e2218523120. https://doi.org/10.1073/pnas.2218523120

    doi:10.1073/pnas.2218523120 2023

  3. [3]

    Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan,A., Shyam, P., Sastry,G.,Askell,A.,Agarwal, S., Herbert-Voss,A., Krueger,G., Henighan,T., Child, R., Ramesh,A., Ziegler,D. M.,Wu, J.,Winter, C.,...Amodei,D.(2020).Languagemodelsarefew-shotlearners.InAdvancesin Neural Information Processing Systems (Vol. 33, pp. 1877–1901...

    arXiv 2020

  4. [4]

    Elderly people and responses to COVID-19 in 27 countries

    Daoust, J.-F.(2020). Elderly people and responses to COVID-19 in 27 countries. PLOSONE,15(7),e0235590.https://doi.org/10.1371/journal.pone.0235590

    work page doi:10.1371/journal.pone.0235590 2020

  5. [6]

    Flaxman,S.,Mishra,S.,Gandy,A.,Unwin,H.J.T.,Mellan,T.A.,Coupland,H., et al. (2020). Estimating the effects of non-pharmaceutical interventions on COVID-19 in Europe. Nature, 584, 257–261. https://doi.org/10.1038/s41586-020-2405-7

    doi:10.1038/s41586-020-2405-7 2020

  6. [7]

    R., McDuff, D., Natarajan, V., Karthikesalingam, A., & Malgaroli, M

    Galatzer-Levy, I. R., McDuff, D., Natarajan, V., Karthikesalingam, A., & Malgaroli, M. (2023). The capability of large language models to measure psychiatric functioning (arXiv preprint arXiv:2308.01834). arXiv. https://arxiv.org/abs/2308.01834

    arXiv 2023

  7. [8]

    J., Chowell, G., et al

    Islam, N., Sharp, S. J., Chowell, G., et al. (2020). Physical distancing interventions and incidence of coronavirus disease 2019: Natural experiment in 149countries.BMJ,370,m2743.https://doi.org/10.1136/bmj.m2743

    work page doi:10.1136/bmj.m2743 2020

  8. [9]

    S., Reid, M., Matsuo, Y., & Iwasawa, Y

    Kojima, T., Gu, S. S., Reid, M., Matsuo, Y., & Iwasawa, Y. (2022). Large language models are zero-shot reasoners (arXiv preprint arXiv:2205.11916). arXiv.https://arxiv.org/abs/2205.11916

    arXiv 2022

Show all 44 references
  1. [10]

    Mitze, T., Kosfeld, R., Rode, J., & Wälde, K. (2020). Face masks considerably reduce COVID-19 cases in Germany. Proceedings of the National Academy of Sciences,117(51),32293–32301.https://doi.org/10.1073/pnas.2015954117

    2020 doi

  2. [11]

    B., Furst, R., Hale, T., Pott, A., & Wood,A.(2021).AworldwideassessmentofchangesinadherencetoCOVID-19 protective behaviours and hypothesized pandemic fatigue

    Petherick, A., Goldszmidt, R., Andrade, E. B., Furst, R., Hale, T., Pott, A., & Wood,A.(2021).AworldwideassessmentofchangesinadherencetoCOVID-19 protective behaviours and hypothesized pandemic fatigue. Nature Human Behaviour,5(9),1145–1160.https://doi.org/10.1038/s41562-021-01181-x

    2021 doi

  3. [12]

    Porat, T., Nyrup, R., Calvo, R.A., Paudyal, P., & Ford, E. (2020). Public health and risk communication during COVID-19: Enhancing psychological needs to promote sustainable behavior change. Frontiers in Public Health, 8, 573397. https://doi.org/10.3389/fpubh.2020.573397

    2020

  4. [13]

    Roda,W.C.,Varughese,M.B.,Han,D.,&Li,M.Y.(2020).Whyisitdifficultto accurately predict the COVID-19 epidemic? Infectious Disease Modelling, 5, 271–281.https://doi.org/10.1016/j.idm.2020.03.001

    2020 doi

  5. [14]

    (2023).Largelanguagemodelscanbeeasilydistractedbyirrelevantcontext.InA

    Shi,F.,Chen,X.,Misra, K.,Yang, J.,Fischer,K.A.,Jimenez,N.A., &Chen, M. (2023).Largelanguagemodelscanbeeasilydistractedbyirrelevantcontext.InA. Krause, E. Brunskill, K. Cho, B. Engelhardt, S. Sabato, & J. Scarlett (Eds.), Proceedingsofthe40thInternationalConferenceonMachineLear...

    2023

  6. [16]

    Šuriņa,S.,Martinsone,K.,Perepjolkina,V.,Kolesnikova,J.,Vainik,U.,Ruža,A., Vrublevska,J.,Smirnova,D.,Fountoulakis,K.N.,&Rancans,E.(2021).Factors relatedtoCOVID-19preventivebehaviors:Astructuralequationmodel.Frontiers inPsychology,12,676521.https://doi.org/10.3389/fpsyg.2021.676521

    2021

  7. [17]

    VanBavel,J.J.,Baicker,K.,Boggio,P.S.,Capraro,V.,Cichocka,A.,Cikara,M., et al. (2020). Using social and behavioural science to support COVID-19 pandemic response. Nature Human Behaviour, 4(5), 460–471. https://doi.org/10.1038/s41562-020-0884-z

    2020 doi

  8. [18]

    J., Caspar, J., Venz, J., et al

    van Lissa, C. J., Caspar, J., Venz, J., et al. (2022). Using machine learning to identifyimportantpredictorsofCOVID-19infectionpreventionbehaviorsduring the early phase of the pandemic. Patterns, 3(4), 100482. https://doi.org/10.1016/j.patter.2022.100482

    2022

  9. [19]

    (2023).Outofone,many:Usinglanguagemodelstosimulatehumansamples

    Argyle,L.P.,Busby,E.C.,Fulda,N.,Gubler,J.R.,Rytting,C.,&Wingate,D. (2023).Outofone,many:Usinglanguagemodelstosimulatehumansamples. PoliticalAnalysis,31(3),337–351.https://doi.org/10.1017/pan.2023.2

    2023 doi

  10. [20]

    https://doi.org/10.48550/arXiv.2408.04667

    Atil,B.,Chittams,A.,Fu,L.,Ture,F.,Xu,L.,&Baldwin,B.(2024).LLM stability:Adetailedanalysiswithsomesurprises.arXiv. https://doi.org/10.48550/arXiv.2408.04667

    2024 doi arXiv

  11. [21]

    Austin,P.C.(2011).Anintroductiontopropensityscoremethodsforreducing theeffectsofconfoundinginobservationalstudies.MultivariateBehavioral Research,46(3),399–424.https://doi.org/10.1080/00273171.2011.568786

    2011

  12. [22]

    Bauer,R.A.(1960).Consumerbehaviorasrisktaking.InR.S.Hancock(Ed.), Dynamicmarketingforachangingworld:Proceedingsofthe43rdConferenceof theAmericanMarketingAssociation(pp.389–398).AmericanMarketing Association

    1960

  13. [23]

    Brewer,N.T.,Chapman,G.B.,Gibbons,F.X.,Gerrard,M.,McCaul,K.D.,& Weinstein,N.D.(2007).Meta-analysisoftherelationshipbetweenrisk perceptionandhealthbehavior:Theexampleofvaccination.HealthPsychology, 26(2),136–145.https://doi.org/10.1037/0278-6133.26.2.136

    2007 doi

  14. [24]

    Larochelle,M.Ranzato,R.Hadsell,M.F.Balcan,&H.Lin(Eds.),Advancesin neuralinformationprocessingsystems(Vol.33,pp.1877–1901).Curran Associates

    Brown,T.B.,Mann,B.,Ryder,N.,Subbiah,M.,Kaplan,J.,Dhariwal,P., Neelakantan,A.,Shyam,P.,Sastry,G.,Askell,A.,Agarwal,S.,Herbert-Voss,A., Krueger,G.,Henighan,T.,Child,R.,Ramesh,A.,Ziegler,D.,Wu,J.,Winter, C.,...Amodei,D.(2020).Languagemodelsarefew-shotlearners.InH. Larochelle,M.Ra...

    2020

  15. [25]

    https://doi.org/10.1016/j.euroecorev.2022.104293

    Chaudhuri,A.,&Howley,P.(2022).TheimpactofCOVID-19vaccinationon mentalwell-being.EuropeanEconomicReview,150,104293. https://doi.org/10.1016/j.euroecorev.2022.104293

    2022

  16. [26]

    Chenchula,N.T.,Chepuri,S.,Soni,A.,Gandhi,D.L.,&Gollapalli,I.S.G. (2023).ClinicalvirologyandeffectofCOVID-19vaccinationandmonoclonal antibodiesagainsthighlyinfectiousSARS-CoV-2OmicronsubvariantBF.7 (BA.5.2.1.7):Asystematicreview.Virology,584,38–43. https://doi.org/10.1016/j.vir...

    2023 doi

  17. [27]

    https://doi.org/10.2196/24341

    deLusignan,S.,LopezBernal,J.,Byford,R.,Amirthalingam,G.,Ferreira,F., Akinyemi,O.,Andrews,N.,Campbell,H.,Dabrera,G.,Deeks,A.,Elliott,A.J., Krajenbrink,E.,Liyanage,H.,McGagh,D.,Okusi,C.,Parimalanathan,V., Ramsay,M.,Smith,G.,Tripathy,M.,...Hobbs,F.D.R.(2021).Influenzaand respirat...

    2021 doi

  18. [28]

    arXiv.https://doi.org/10.48550/arXiv.2402.10811

    Hu,T.,&Collier,N.(2024).QuantifyingthepersonaeffectinLLMsimulations. arXiv.https://doi.org/10.48550/arXiv.2402.10811

    2024 doi arXiv

  19. [30]

    https://doi.org/10.1056/NEJMoa2001316

    Li,Q.,Guan,X.,Wu,P.,Wang,X.,Zhou,L.,Tong,Y.,Ren,R.,Leung,K.S.M., Lau,E.H.Y.,Wong,J.Y.,Xing,X.,Xiang,N.,Wu,Y.,Li,C.,Chen,Q.,Li,D., Liu,T.,Zhao,J.,Liu,M.,...Feng,Z.(2020).Earlytransmissiondynamicsin Wuhan,China,ofnovelcoronavirus-infectedpneumonia.TheNewEngland JournalofMedicine...

    2020 doi

  20. [31]

    https://doi.org/10.1145/3560815

    Liu,P.,Yuan,W.,Fu,J.,Jiang,Z.,Hayashi,H.,&Neubig,G.(2023).Pre-train, prompt,andpredict:Asystematicsurveyofpromptingmethodsinnatural languageprocessing.ACMComputingSurveys,55(9),Article195. https://doi.org/10.1145/3560815

    2023 doi

  21. [32]

    JournaloftheAmericanStatisticalAssociation,46(253),68–78

    Massey,F.J.,Jr.(1951).TheKolmogorov–Smirnovtestforgoodnessoffit. JournaloftheAmericanStatisticalAssociation,46(253),68–78. https://doi.org/10.1080/01621459.1951.10500769

    1951

  22. [33]

    11048–11064).AssociationforComputationalLinguistics

    Min,S.,Lewis,M.,Hajishirzi,H.,&Zettlemoyer,L.(2022).Rethinkingtherole ofdemonstrations:Whatmakesin-contextlearningwork?InProceedingsofthe 2022ConferenceonEmpiricalMethodsinNaturalLanguageProcessing(pp. 11048–11064).AssociationforComputationalLinguistics. https://doi.org/10.186...

    2022 doi

  23. [34]

    https://doi.org/10.1093/biomet/70.1.41

    Rosenbaum,P.R.,&Rubin,D.B.(1983).Thecentralroleofthepropensity scoreinobservationalstudiesforcausaleffects.Biometrika,70(1),41–55. https://doi.org/10.1093/biomet/70.1.41

    1983 doi

  24. [35]

    Savadori,L.,&Lauriola,M.(2021).Riskperceptionandprotectivebehaviors duringtheriseoftheCOVID-19outbreakinItaly.FrontiersinPsychology,11, 577331.https://doi.org/10.3389/fpsyg.2020.577331

    2021

  25. [36]

    (2013).Determiningstatisticalsignificance(alpha)andreportingstatisticaltrends: Controversies,issues,andfacts.ComprehensivePsychology,2

    Schumm,W.R.,Pratt,C.C.,Hartenstein,J.L.,Jenkins,B.,&Johnson,D. (2013).Determiningstatisticalsignificance(alpha)andreportingstatisticaltrends: Controversies,issues,andfacts.ComprehensivePsychology,2. https://doi.org/10.2466/03.CP.2.10

    2013 doi

  26. [37]

    https://doi.org/10.1126/science.3563507

    Slovic,P.(1987).Perceptionofrisk.Science,236(4799),280–285. https://doi.org/10.1126/science.3563507

    1987 doi

  27. [38]

    Xia,Q.,Yang,Y.,Wang,F.,Huang,Z.,Qiu,W.,&Mao,A.(2024).Casefatality ratesofCOVID-19duringepidemicperiodsofvariantsofconcern:A meta-analysisbycontinents.InternationalJournalofInfectiousDiseases,141, 106950.https://doi.org/10.1016/j.ijid.2024.01.017

    2024 doi

  28. [39]

    Perception-based influence factorsof intention to adopt COVID-19 epidemic prevention in China

    Ahmad, M.,Iram,K., & Jabeen, G.(2020). Perception-based influence factorsof intention to adopt COVID-19 epidemic prevention in China. Environmental Research,190,109995.https://doi.org/10.1016/j.envres.2020.109995

    2020

  29. [40]

    Cui, H., Chen, B., Jiang, Y., Tao, Y., Zhu, X., & Cai, Z. (2021). Toxicity of 17 disinfection by-products to different trophic levels of aquatic organisms: Ecological risks and mechanisms. Environmental Science & Technology, 55(15), 10534–10541.https://doi.org/10.1021/acs.est.0c08796

    2021 doi

  30. [41]

    Díaz, R., & Cova, F. (2022). Reactance, morality, and disgust: The relationship between affective dispositions and compliance with official health recommendations during the COVID-19 pandemic. Cognition & Emotion, 36(1), 120–136.https://doi.org/10.1080/02699931.2021.1941783

    2022

  31. [42]

    A., & Klein, W

    Ferrer, R. A., & Klein, W. M. P. (2015). Risk perceptions and health behavior. Current Opinion in Psychology, 5, 85–89. https://doi.org/10.1016/j.copsyc.2015.03.012

    2015 doi

  32. [43]

    Hu, L., Liu, Z., Zhao, Z., Hou, L., Nie, L., & Li, J. (2024). A survey of knowledge enhanced pre-trained language models. IEEE Transactions on Knowledge and Data Engineering, 36(4), 1413–1430. https://doi.org/10.1109/TKDE.2023.3310002

    2024

  33. [44]

    Li, Z., Zhang, Q., Wang, C., Zhang, F., Liu, Y., Zhang, L., & Xie, D. (2025). Balancing act: How community measures shape environmental risk perceptions and preventive behaviours during the COVID-19 pandemic. Environmental Impact Assessment Review, 115, 108062. https://doi.org...

    2025

  34. [45]

    Mertens, G., Lodder, P., Smeets, T., & Duijndam, S. (2023). Pandemic panic? Results of a 14-month longitudinal study on fear of COVID-19. Journal of AffectiveDisorders,322,15–23.https://doi.org/10.1016/j.jad.2022.11.008

    2023 doi

  35. [47]

    Wong, C. M. L., & Jensen, O. (2020). The paradox of trust: Perceived risk and publiccompliance duringthe COVID-19 pandemic inSingapore.Journal ofRisk Research,23(7–8),1021–1030.https://doi.org/10.1080/13669877.2020.1756386

    2020

  36. [48]

    (2025).Applying large language modelstotravel satisfaction analysis.arXiv.https://doi.org/10.48550/arXiv.2505.23262

    Xu,P.,&Wang, D. (2025).Applying large language modelstotravel satisfaction analysis.arXiv.https://doi.org/10.48550/arXiv.2505.23262

    2025 doi

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.