Pith. sign in

REVIEW 3 major objections 8 minor 33 references

Tackling One Health Risks: How Large Language Models are leveraged for Risk Negotiation and Consensus-building

T0 review · 3 major / 8 minor · reviewed 2026-08-04 · 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 human-supervised workflow of LLM-based negotiating agents can help multi-sector One Health stakeholders reach consensus on contested risk-management choices, and it supports this with two proof-of-concept case scena

desk verdict Reproducible LLM-assisted negotiation workflow for One Health, honestly framed as a proof-of-concept; role-play evaluation limits the consensus claims. read the letter →

arxiv 2509.09906 v1 pith:IXBB2IAL submitted 2025-09-12 cs.MA cs.AI

Alexandra Fetsch , Iurii Savvateev , Racem Ben Romdhane , Martin Wiedmann , Artemiy Dimov , Maciej Durkalec , Josef Teichmann , Jakob Zinsstag
show 7 more authors
Konstantinos Koutsoumanis Andreja Rajkovic Jason Mann Mauro Tonolla Monika Ehling-Schulz Matthias Filter Sophia Johler
This is my paper
classification cs.MAcs.AI
keywords OneHealthriskanalysisnegotiationlargelanguagemodelsmulti-agentsystemshuman-in-the-loopconsensus-buildingproofofconcept
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 is trying to establish that large language models can make multi-stakeholder risk negotiation in One Health workable under real-world time and information constraints. It proposes a human-supervised workflow in which LLM-based agents, each representing a stakeholder, simulate a cooperative negotiation over risk-management options; the resulting deal proposals are then discussed and adjusted by the human stakeholders. The authors tested this in two cases—whether to keep using a biopesticide and whether to restrict wild boar hunting—and report that the role-playing stakeholder groups reached consensus within the two-hour sessions. If this holds, it gives resource-limited organizations an open-source way to support cross-sectoral decisions that currently lack structured negotiation tools.

What carries the argument

The engine is a human-supervised multi-agent negotiation game. Each LLM-based agent is prompted with a stakeholder's position paper, the agreed issue/option list, confidential preference scores, and game rules; rounds proceed with agents endorsing existing deals or proposing new ones, and a moderator's opening suggestion shifts which deals emerge. A post-analysis layer draws each proposed deal as a line over preference surfaces and exposes the 'scratchpad' rationale behind proposals, so humans can see exactly which issue blocked agreement. The load-bearing device is the 100-point scoring template: it converts qualitative positions into numbers that agents can optimize over while keeping indi

What would settle it

Run the same two scenarios with authentic stakeholder representatives—regulators, farmers, hunters, and animal-welfare advocates—under the same two-hour constraint and check whether they reach a deal and whether that deal falls inside the simulated deal distribution. If such groups reject all machine-suggested deals or fail to converge, the central claim would not transfer outside the role-play setting.

Watch

Extended reading notes

Core claim

The authors claim to have operationalized negotiation-centered risk analysis by combining an LLM-based multi-agent negotiation game with a human-in-the-loop review stage. In their procedure, each stakeholder first writes a position paper, then the group agrees on a fixed list of issues and options, and each stakeholder privately assigns a 100-point budget across issues and options to express importance and flexibility. These confidential scores are fed into agents that negotiate a non-zero-sum game over multiple rounds, producing a distribution of proposed 'deals'—each a package picking one option per issue. Stakeholders then inspect the simulated deals and their rationales, may choose to di

Load-bearing premise

The proof-of-concept depends on project team members role-playing the stakeholders; if real stakeholders with power asymmetries, veto rights, and entrenched interests negotiate differently, the observed two-hour consensuses do not establish the framework's usefulness.

Editorial extensions

If this is right

  • Groups can move from position papers to a concrete deal package within a two-hour session, compressing problem formulation, valuation, and negotiation into one exercise.
  • Because the implementation is open source, web-based, and not tied to a particular LLM, organizations with limited AI resources can adapt it to their own risk topics.
  • The same pipeline can be run without human discussion to pre-explore possible negotiation outcomes, serving as a rehearsal for real round-tables.
  • Controlled disclosure of partial scores—revealing only the issues one cares about most—can unlock compromises that fully secret preferences would block.
  • The moderator's identity measurably changes which deals are proposed, so facilitator selection is a substantive design choice, not an administrative detail.

Reading between the lines

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

  • A fair next test would compare this pipeline against conventional facilitated round-tables using real stakeholder groups, measuring time-to-agreement, satisfaction, and whether agreements hold after the session.
  • Because the human-approved final deals can diverge from the simulated equilibrium, the framework is best read as deliberation support for compromise discovery rather than equilibrium computation; quantifying that divergence would clarify what the simulation actually predicts.
  • The role-played stakeholder design means the reported consensus is a usability proof, not evidence about how real power asymmetries, vetoes, and entrenched interests would play out; trials with actual regulators, farmers, hunters, and advocates would be the natural next step.
  • The same multi-agent setup could be extended to adversarial or bad-faith negotiation scenarios, letting groups stress-test their consensus against sabotage and strategic misrepresentation before real talks.
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

3 major / 8 minor

Summary. The paper presents an AI-assisted negotiation framework for One Health risk analysis, combining LLM-based agents with a human-in-the-loop (HIL) approach. The workflow operationalizes a previously proposed six-step negotiation-centered risk analysis process, with a focus on steps (iii) risk assessment/valuation and (iv) risk negotiation. Proof-of-concept implementations are described for two scenarios: use of Bacillus thuringiensis as a biopesticide and wild boar population control. In each, project team members role-played three stakeholder groups, provided position papers and confidential preference scores, and used LLM-generated issue/option lists and simulated deal distributions to negotiate a final consensus. The authors report that in both scenarios consensus was reached within the two-hour time constraint. The paper also provides open-source code, a Zenodo repository with templates and results, and detailed supplementary protocols.

Significance. If the framework is validated, it would offer a reproducible, open-source tool for structuring multi-stakeholder One Health negotiations under time constraints. The paper makes several contributions: a concrete step-by-step pipeline linking LLM-based multi-agent simulation to a defined human oversight process; two detailed, realistic One Health case scenarios; and public release of code, templates, and data. The strongest strength is the explicit procedural formalization of steps 3a-3e and step 4, which others could adopt or adapt. However, the current evidence is proof-of-concept only: the 'successful consensus' outcome is based on role-play by project team members, with no baseline, no quantitative outcome measure, and no external validation. The significance of the claimed results is therefore contingent on future validation with real stakeholder groups.

major comments (3)
  1. [Methods, 'Case scenarios and practical approach'] The paper's central demonstration—that stakeholders successfully completed risk negotiation—rests on exercises where 'project team members assumed roles representing one of three stakeholder groups.' Participants are co-authors/experts with a vested interest in the project's success, which is a selection bias. Real multi-sectoral stakeholders with asymmetric power, veto rights, and entrenched institutional mandates may behave differently. The Discussion generalizes without this caveat, stating 'in both of our case scenarios the stakeholders were able to successfully complete the risk negotiation,' and the abstract extends to 'stakeholders.' Please either temper the claims to a role-play proof-of-concept or provide external validation with actual stakeholders.
  2. [Abstract and Discussion (claims of mitigation and consensus)] The abstract claims the framework 'mitigates information overload and augments decision-making process under time constraints,' and the Discussion states that stakeholders 'successfully complete the risk negotiation within the time-constraints requirement.' No baseline, control condition, or quantitative outcome is reported. The only measured outcome is self-reported acceptance by the participants themselves. Concrete metrics are needed—e.g., time to consensus, number of rounds, agreement scores, satisfaction, or comparison with an unaided manual negotiation—or the claims must be restricted to 'the pipeline ran end-to-end in two simulated scenarios.'
  3. [Methods, steps (iiie) and (iv); Figure 2] The same individuals who supply the confidential preference scores (step iiie) are the ones who discuss and approve the simulated deals (step iv), and those scores are directly used to prompt the LLM agents. The 'suggested equilibrium' is therefore a function of the very inputs used to validate it. This is not a fatal flaw for a decision-support tool, but it means the observed consensus cannot be interpreted as an independent validation of the LLM-based negotiation. Please explicitly frame the results as preference aggregation followed by human discussion, and separate any claims about the LLM's negotiation ability from claims about the overall workflow's usefulness.
minor comments (8)
  1. [Results, step (iv)] Typo: 'equillibrium' should be 'equilibrium.'
  2. [Affiliation 7] Typo: 'Insitute' should be 'Institute.'
  3. [References, ref. 11] Typo: 'Higgings' should likely be 'Higgins.'
  4. [Supplementary Text S3, scoring guide] In the quick guide, 'areas where a comprise is feasible' should be 'compromise.'
  5. [Supplementary Text S6] Formatting issue: 'issue C was give n highest priority' has a stray space; please correct.
  6. [Methods, step 4 description] 'non-zero game' should be 'non-zero-sum game' to match standard terminology.
  7. [Figure 2 caption] The term 'Nash equilibrium' is used loosely for a cooperative negotiation game; consider clarifying that the model searches for a compromise point rather than a formal Nash equilibrium, or define the term as used here.
  8. [Supplementary Text S2, step 4] The claim of reproducibility from 'multiple iterations' would be strengthened by reporting random seeds or variance across runs; Fig. S1-S4 show distributions but no statistical summary.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the workflow is a facilitated negotiation demonstration, not a fitted prediction; the self-cited framework is not load-bearing.

full rationale

The paper does not claim to derive an empirical prediction from fitted parameters. The LLM negotiation simulation uses stakeholder preference scores as inputs, but the paper explicitly treats the simulated deals as suggestions for human discussion ('stakeholders were provided with a report containing the distribution of the most popular deals, and were asked to discuss them to determine whether a compromise can be reached'). The human-in-the-loop step can and did reject or modify the simulated deals: in case scenario 1 the consumer representative rejected the most approved combination and the final agreement (A1 B2 C2 D2 E1 F1) differed from the most approved simulated deal (A1 B2 C2 D2 E1 F2); in case scenario 2 the final package also combined elements beyond the two most-approved lists. Thus the final consensus is not forced by construction from the input scores; failure was possible. The only notable self-citation is Ehling-Schulz et al. (2024), which supplies the conceptual six-step framework, but the present contribution is the operationalization with LLM agents, and no load-bearing claim reduces to that citation. The use of project-team members as role-playing stakeholders is a real external-validity limitation, but it is a threat to generalization, not a circularity in the derivation chain.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The framework introduces no new physical or mathematical entities. Its evidential weight rests entirely on procedural assumptions about LLM fidelity, the representativeness of role-played stakeholders, and the appropriateness of the underlying negotiation game plus the Nash equilibrium concept. These assumptions are untested in the paper.

assumptions (4)
  • domain assumption LLM agents prompted with position papers and preference scores produce negotiation behavior representative enough of real stakeholder deliberation.
    Invoked in Methods and Discussion step 4, where the simulated negotiation is treated as a basis for human consensus and the paper argues suggestions are 'semantically enriched'. The limitation section acknowledges LLM hallucinations and biases, which is why HIL is used, but no fidelity check is performed.
  • domain assumption The cooperative negotiation game from Abdelnabi et al. (2023) is a valid model of multi-party risk negotiation.
    The pipeline is built directly on this framework (Methods: 'using a cooperative scenario developed by Abdelnabi and colleagues (18)'); no independent validation of the game's representativeness is provided.
  • domain assumption Project team members role-playing farmer, consumer, food safety authority, hunter, and animal protection representatives generate valid evidence about the framework's usefulness in real settings.
    Methods states 'For practicality, project team members assumed roles representing one of three stakeholder groups in each scenario.' This is the basis for the demonstration and is a major generalization risk.
  • domain assumption Nash equilibrium is the appropriate notion of compromise for multi-issue scoring negotiation.
    Figure 2A and step 4 describe the negotiation as a search for a Nash equilibrium; no justification is given for this equilibrium concept over alternatives.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Tackling One Health Risks: How Large Language Models are leveraged for Risk Negotiation and Consensus-building." pith.science (2026). https://pith.science/paper/IXBB2IAL

@misc{pith2026250909906,
  author       = {Pith},
  title        = {Pith review of: Tackling One Health Risks: How Large Language Models are leveraged for Risk Negotiation and Consensus-building},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IXBB2IAL}},
  note         = {Machine review of arXiv:2509.09906}
}
read the original abstract

Key global challenges of our times are characterized by complex interdependencies and can only be effectively addressed through an integrated, participatory effort. Conventional risk analysis frameworks often reduce complexity to ensure manageability, creating silos that hinder comprehensive solutions. A fundamental shift towards holistic strategies is essential to enable effective negotiations between different sectors and to balance the competing interests of stakeholders. However, achieving this balance is often hindered by limited time, vast amounts of information, and the complexity of integrating diverse perspectives. This study presents an AI-assisted negotiation framework that incorporates large language models (LLMs) and AI-based autonomous agents into a negotiation-centered risk analysis workflow. The framework enables stakeholders to simulate negotiations, systematically model dynamics, anticipate compromises, and evaluate solution impacts. By leveraging LLMs' semantic analysis capabilities we could mitigate information overload and augment decision-making process under time constraints. Proof-of-concept implementations were conducted in two real-world scenarios: (i) prudent use of a biopesticide, and (ii) targeted wild animal population control. Our work demonstrates the potential of AI-assisted negotiation to address the current lack of tools for cross-sectoral engagement. Importantly, the solution's open source, web based design, suits for application by a broader audience with limited resources and enables users to tailor and develop it for their own needs.

Figures

Figures reproduced from arXiv: 2509.09906 by the authors.

Figure 1
Figure 1. Workflow of negotiation-centered risk analysis (incorporating the visualization of the risk negotiation framework of Ehling-Schulz, et al. (1). The figure visualizes the negotiation-centered risk analysis framework incorporating 6 different steps as follows (i) stakeholder round table establishment; (ii) problem formulation; (iii) risk assessment and valuation; (iv) risk negotiation; (v) communication and implementa… view at source ↗
Figure 4
Figure 4. Integrating agent-based modeling in the effect dimensions specification. Agent-based modeling was used to facilitate a preliminary discussion about the raised problem and was used for issue/option definition as described in the corresponding section of [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

33 extracted references · 1 linked inside Pith

  1. [2]

    Strict Safety Measures

    Follow the 2-step scoring approach: First, for each issue define the maximum score (points) you put on this issue as follows:  Max score weights the importance of the issue to you  You can put scores from 0 to 100: o 0- the lowest importance o 100- the highest importance  sum of scores of all issues should be 100 in total (remember: you have 100 points...

  2. [3]

    Bacillus cereus—a multifaceted opportunistic pathogen

    Messelhäußer U, Ehling-Schulz M. Bacillus cereus—a multifaceted opportunistic pathogen. Current Clinical Microbiology Reports. 2018;5(2):120-5

    2018

  3. [4]

    Occurrence of natural Bacillus thuringiensis contaminants and residues of Bacillus thuringiensis-based insecticides on fresh fruits and vegetables

    Frederiksen K, Rosenquist H, Jørgensen K, Wilcks A. Occurrence of natural Bacillus thuringiensis contaminants and residues of Bacillus thuringiensis-based insecticides on fresh fruits and vegetables. Appl Environ Microbiol. 2006;72(5):3435-40

    2006

  4. [5]

    The higher the value, the higher the preference

    General rules You have 100 points that can be used to define your preferences. The higher the value, the higher the preference. You do not know the scores of the other parties, but you do know the descriptions of the other parties, so you can share your guesses on what other may think (optional). Your goal is to define scores reflecting your preferences, ...

  5. [6]

    Strict Control

    Follow the 2-step scoring approach: First, for each issue define the maximum score (points) you put on this issue as follows:  Max score weights the importance of the issue to you  You can put scores from 0 to 100: o 0- the lowest importance o 100- the highest importance  sum of scores of all issues should be 100 in total (remember: you have 100 points...

  6. [7]

    Bacillus thuringiensis: a successful insecticide with new environmental features and tidings

    Jouzani GS, Valijanian E, Sharafi R. Bacillus thuringiensis: a successful insecticide with new environmental features and tidings. Appl Microbiol Biotechnol. 2017;101(7):2691-711. Epub 20170224

    2017

  7. [8]

    Pest control and resistance management through release of insects carrying a male-selecting transgene

    Harvey-Samuel T, Morrison NI, Walker AS, Marubbi T, Yao J, Collins HL, et al. Pest control and resistance management through release of insects carrying a male-selecting transgene. BMC Biology. 2015;13(1):49

    2015

  8. [9]

    Effective alternatives to hunting exist to tackle disease spread, while ensuring animal welfare

    Anonymous. Effective alternatives to hunting exist to tackle disease spread, while ensuring animal welfare. EuroGroupForAnimals; 2022 [2024-10-22]; Available from: https://www.eurogroupforanimals.org/news/effective-alternatives-hunting-exist-tackle-disease- spread-while-ensuring-animal-welfare

    2022

Show all 33 references
  1. [10]

    Anonymous. Shooting and trapping in the forest have nothing to do with animal welfare - How animals suffer during hunting.: Deutscher Tierschutzbund; 2024 [2024-10-22]; Available from: https://www.tierschutzbund.de/en/animals-topics/wild-animals/hunting

    2024

  2. [11]

    Common occurrence of enterotoxin genes and enterotoxicity in Bacillus thuringiensis

    Gaviria Rivera AM, Granum PE, Priest FG. Common occurrence of enterotoxin genes and enterotoxicity in Bacillus thuringiensis. FEMS Microbiol Lett. 2000;190(1):151-5

    2000

  3. [12]

    Isolation and characterization of Bacillus cereus-like bacteria from faecal samples from greenhouse workers who are using Bacillus thuringiensis-based insecticides

    Jensen GB, Larsen P, Jacobsen BL, Madsen B, Wilcks A, Smidt L, et al. Isolation and characterization of Bacillus cereus-like bacteria from faecal samples from greenhouse workers who are using Bacillus thuringiensis-based insecticides. International Archives of Occupational and...

    2002

  4. [14]

    Enterotoxin production of Bacillus thuringiensis isolates from biopesticides, foods, and outbreaks

    Johler S, Kalbhenn EM, Heini N, Brodmann P, Gautsch S, Ba ğcioğlu M, et al. Enterotoxin production of Bacillus thuringiensis isolates from biopesticides, foods, and outbreaks. Frontiers in Microbiology. 2018;9

    2018

  5. [15]

    Enteropathogenic potential of Bacillus thuringiensis isolates from soil, animals, food and biopesticides

    Schwenk V, Riegg J, Lacroix M, Märtlbauer E, Jessberger N. Enteropathogenic potential of Bacillus thuringiensis isolates from soil, animals, food and biopesticides. Foods. 2020;9(10):1484

    2020

  6. [16]

    To o many wild boar? Modelling fertility control and culling to reduce wild boar numbers in isolated populations

    Croft S, Franzetti B, Gill R, Massei G. To o many wild boar? Modelling fertility control and culling to reduce wild boar numbers in isolated populations. PLoS One. 2020;15(9):e0238429. Epub 20200918

    2020

  7. [17]

    Gieser T. Hunting wild animals in Germany: Conflicts between wildlife management and ‘traditional’ practices of Hege, in: Michaela Fenske, Bernhard Tschofen (eds) Managing the Return of the Wild Human Encounters with Wolves in Europe. London: Routledge, pp. 164-179

  8. [18]

    Wild boar populations up, numbers of hunters down? A review of trends and implications for Europe

    Massei G, Kindberg J, Licoppe A, Ga čić D, Šprem N, Kamler J, et al. Wild boar populations up, numbers of hunters down? A review of trends and implications for Europe. Pest Manag Sci. 2015;71(4):492-500. Epub 20150129

    2015

  9. [19]

    Efficacy of hunting, feeding, and fencing to reduce crop damage by wild boars

    Geisser H, Reyer H-U. Efficacy of hunting, feeding, and fencing to reduce crop damage by wild boars. The Journal of Wildlife Management. 2004;68(4):939-46

    2004

  10. [20]

    Impact of w ild boar (Sus scrofa) in its introduced and native range: a review

    Barrios-Garcia MN, Ballari SA. Impact of w ild boar (Sus scrofa) in its introduced and native range: a review. Biological Invasions. 2012;14(11):2283-300

    2012

  11. [21]

    African Swine Fever: Transmission, spread, and control through biosecurity and disinfection, including polish trends

    Juszkiewicz M, Walczak M, Wo źniakowski G, Podgórska K. African Swine Fever: Transmission, spread, and control through biosecurity and disinfection, including polish trends. Viruses. 2023;15(11). Epub 20231119

    2023

  12. [22]

    Risk negotiation: a framework for One Health risk analysis

    Ehling-Schulz M, Filter M, Zinsstag J, Koutsoumanis K, Ellouze M, Teichmann J, et al. Risk negotiation: a framework for One Health risk analysis. Bull World Health Organ. 2024;102(6):453-6. Epub 20240508

    2024

  13. [23]

    Low-cost electric fencing for peaceful coexistence: An analysis of human-wildlife conflict mitigation strategies in smallholder agriculture

    Feuerbacher A, Lippert C, Kuenzang J, Subedi K. Low-cost electric fencing for peaceful coexistence: An analysis of human-wildlife conflict mitigation strategies in smallholder agriculture. Biological Conservation. 2021;255:108919

    2021

  14. [24]

    Efficiency of spreading maize in the garrigues to reduce wild boar (Sus scrofa) damage to Mediterranean vineyards

    Calenge C, Maillard D, Fournier P, Fouque C. Efficiency of spreading maize in the garrigues to reduce wild boar (Sus scrofa) damage to Mediterranean vineyards. European Journal of Wildlife Research. 2004;50(3):112-20

    2004

  15. [25]

    www.ArXiv.com

    ArXiv. www.ArXiv.com. 20

  16. [26]

    www.wikipedia.org

    Wikipedia. www.wikipedia.org

  17. [27]

    www.duckduckgo.com

    Duckduckgo. www.duckduckgo.com

  18. [28]

    https://ollama.com/

    Ollama. https://ollama.com/

  19. [29]

    https://openai.com

    OpenAI. https://openai.com

  20. [30]

    https://gemini.google.com/

    Alphabet. https://gemini.google.com/

  21. [31]

    https://huggingface.co/

    HuggingFace. https://huggingface.co/

  22. [32]

    https://www.llama.com/

    Llama. https://www.llama.com/

  23. [33]

    https://www.nvidia.com/de-de/

    Nvidia. https://www.nvidia.com/de-de/

  24. [35]

    Cooperation, competition, and maliciousness: LLM-stakeholders interactive negotiation

    Abdelnabi S, Gomaa A, Sivaprasad S, Schönherr L, Fritz M. Cooperation, competition, and maliciousness: LLM-stakeholders interactive negotiation. arXiv e-prints. 2023:arXiv:2309.17234

    2023 arXiv

  25. [38]

    Should B. thuringiensis continue to be used as biopesticide in Central Europe?

    Tessler MH, Bakker MA, Jarrett D, Sheahan H, Chadwick MJ, Koster R, et al. AI can help humans find common ground in democratic deliberation. Science. 2024;386(6719):eadq2852. 39. Baarslag T, Hendrikx MJC, Hindriks KV, Jonker CM. Learning about the opponent in automated bilater...

    2024

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.