CoPlan: A Trustworthy Co-Intelligence Interface for Care Planning through Role-Based Contestable Argument Graphs

H\'el\`ene Fournier; Hung Cao; Hung Truong Thanh Nguyen; Makoto Itoh; Piper Jackson; Rene Richard; Shannon Freeman

REVIEW 5 major objections 5 minor 28 references

CoPlan: A Trustworthy Co-Intelligence Interface for Care Planning through Role-Based Contestable Argument Graphs

T0 review · 5 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read CoPlan argues that AI-generated care recommendations should be reviewable proposals, not fixed decisions, and demonstrates an interface where human care planners edit the supporting arguments before a final plan is produced.

desk verdict A coherent design proposal for contestable care planning with a real interface and clear workflow, but no evaluation and an unverified link from edited arguments to the final plan. read the letter →

arxiv 2608.05107 v1 pith:XBCNT6TS submitted 2026-08-05 cs.AI cs.MAcs.SE

Hung Truong Thanh Nguyen , Hélène Fournier , Piper Jackson , Makoto Itoh , Shannon Freeman , Rene Richard , Hung Cao This is my paper

classification cs.AIcs.MAcs.SE

keywords careplanningcontestableAIhuman-AIcollaborationmulti-agentsystemsargumentationquantitativebipolaraginginplacehuman-in-the-loop

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

This paper presents CoPlan, an interface for human-AI care planning built on two ideas: co-intelligence and contestability. The central claim is that AI recommendations in care planning should be treated as reviewable proposals rather than fixed outputs, so that clinicians, patients, and caregivers can inspect, challenge, and revise the reasoning behind them. CoPlan organizes that reasoning as role-based argument graphs: specialized AI agents propose candidate interventions and generate supporting and challenging arguments for each, while human care planners accept, reject, modify, or add arguments before the system assembles the final plan. The paper demonstrates the pipeline in an aging-in-place scenario, showing how expert roles, evidence, and human edits flow into a prioritized care plan. A sympathetic reader would care because the design directly targets the gap between explainable AI, which shows why a recommendation was made, and contestable AI, which lets people change it.

What carries the argument

The central object is the role-based contestable argument graph, combined with a quantitative bipolar argumentation framework (QBAF). In the QBAF $Q=\langle X, R^+, R^-, \tau\rangle$, each argument $x$ has an intrinsic strength $\tau(x)\in[0,1]$, support and challenge relations $R^+$ and $R^-$ carry influence, and the update equation $f(x)=\sigma(\tau(x)+I(x,f))$ iterates to convergence, where $I(x,f)$ sums supportive influence and subtracts challenging influence. Option-level scores $F(o_i)$ aggregate the degrees of supporting and challenging arguments. This machinery makes AI reasoning visible, editable, and recomputable: when a human rejects an argument or adds a new one, the updated set $\Gamma_H$ is re-scored, so human judgment propagates through the graph and into the final plan's priorities.

What would settle it

A concrete test would be to take a set of care cases with known expert consensus, run CoPlan's scoring and aggregation, and compare the resulting option priorities against the consensus; if the LLM validity scores disagree with human experts on which intervention should rank first in a substantial fraction of cases, the ordering mechanism fails.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that care planning can be structured as a contestable co-intelligence workflow rather than an automated optimization. CoPlan formalizes the environment as a tuple $M=\langle P, D, O, A, \Gamma, H, V, \Pi\rangle$ and runs four stages: complexity assessment and adaptive care team recruitment, option generation and role-based argumentation, human-in-the-loop contestation, and final plan generation with human approval. The central step is Stage 3: every argument carries its originating role and its strength scores, the human care team edits the full argument set, and then the quantitative bipolar argumentation framework recomputes degrees from support and challenge relations. The final care plan priorities are determined by these human-validated argument scores, so the plan is the output of accountable negotiation rather than raw model inference.

Load-bearing premise

The whole ranking rests on the assumption that the AI-assigned intrinsic strength scores and the post-review validity scores from the Argument Validation prompt are accurate enough to order interventions, since the paper gives no external validation of these scores and any error flows into final plan priorities.

Editorial extensions

If this is right

If CoPlan's approach is correct, AI care-planning systems can preserve human authority by making argument graphs the unit of review rather than final outputs.
Human edits to arguments are not cosmetic: they change option-level scores and therefore change the priority of final recommendations.
Adaptive team recruitment ties the number and mix of expert roles to case complexity, so more complex cases get broader interdisciplinary review.
Because each argument is tagged by role and grounded in retrieved evidence, the final plan can cite specific evidence and show which professional viewpoint produced each point.
The scheduling agent can turn approved plan components into concrete appointments, connecting contestable planning to practical follow-through.

Reading between the lines

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

Beyond the paper, the same role-based contestable argument graph could be applied to other high-stakes human-AI decisions where multiple professional viewpoints matter, such as hospital discharge planning, disability accommodation assessment, or social-services eligibility.
A testable extension would be a controlled study comparing plans produced with and without the contestation stage, measuring whether human edits actually shift final priorities and whether participants report greater ownership of the plan.
The paper does not quantify how much review time contestation adds; a practical extension would instrument the interface to measure the cost of editing versus the value in changed outcomes.
Because the final rankings depend on LLM-generated validity scores, a natural stress test is to perturb those scores slightly and check whether the option-level ordering remains stable.

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.

Desk Editor's Note

read the letter

CoPlan is a well-organized design paper for a contestable care-planning interface. What is actually new is the integration: role-tagged arguments from specialized LLM agents, a QBAF layer that computes argument and option scores, and a human-editable argument graph sitting before final plan generation. The underlying pieces (multi-agent LLMs, QBAF, contestability) are established, so this is an application and extension, not a breakthrough. That is fine; the packaging is sensible and the implementation is public, which deserves credit.

The workflow is clear and concrete. The prompts are specific, the interRAI grounding is a good idea, and the aging-in-place scenario is plausible. Treating AI recommendations as reviewable proposals rather than fixed outputs is a genuinely useful framing for care planning, and the role-based review interface gives stakeholders something they can actually interact with.

The soft spots are real and mostly about evidence. There is no evaluation: no user study, no baseline, no quantitative metric. The central claim about preserving human agency and accountability rests on a screenshot. More specifically, the final care plan operator is an LLM prompt that says 'consider the strength of arguments,' but there is no verification that the QBAF scores, or even human edits, causally change the final plan. The stress-test is right that the scores could be ignored. It overstates the failure, though: when a human rejects or edits an argument, that changes the text of the validated arguments fed to the final prompt, so the plan is at least a function of what the human kept. That is a weaker guarantee than 'scores determine priority,' but it is not performative. The paper should state this weaker guarantee honestly, or add an ablation showing how edits change outputs.

Two smaller things. The aggregation function g and the influence weights alpha and beta are left unspecified; for a design paper that is minor, but if the QBAF is meant to matter, the authors should commit to at least one instantiation and test its effect. And the stated limitation — that validity scores support review rather than autonomous decision-making — is honest but does not address the score-to-plan link.

Who this is for: people working on contestable AI, human-AI decision support, and care-planning interfaces. It is not a clinical validation and should not be read as one. But the design direction is worth engaging seriously, and a good referee could push the authors to add a small user study or an outcome-based ablation. I would send it to peer review, expecting significant revision rather than acceptance as is.

Referee Report

5 major / 5 minor

Summary. The paper presents CoPlan, a multi-agent interface for aging-in-place care planning. The system uses role-based LLM agents to generate candidate interventions and supporting/challenging arguments, presents these as an editable argument graph, lets human care planners accept, reject, modify, or add arguments, re-computes argument strengths via a quantitative bipolar argumentation framework (QBAF), and produces a final care plan through a separate LLM prompt. The authors claim that this design treats AI recommendations as reviewable proposals and thereby preserves human agency and clinical accountability. A single illustrative aging-in-place scenario is presented as a demonstration, along with an open-source repository and video link.

Significance. If the central claim were substantiated, CoPlan would be a useful design exemplar for contestable AI in care planning, showing how argumentation-based interfaces can operationalize human oversight. The paper also usefully frames co-intelligence and contestability as complementary principles. The authors explicitly release an implementation, which is a strength. However, the paper currently provides no evaluation: no user study, no metrics, no verification that human edits change the final plan, and no external validation of the LLM-generated scores. The significance therefore rests on the design proposal rather than on demonstrated effectiveness.

major comments (5)

[Stage 4 (Section 3.2) and Conclusion (Section 5)] The central claim that human contestation causally affects the final care plan is not supported. Section 3.2 says the care plan operator Π uses the updated option-level scores F(o_i) to determine priority, but the implementation is a single prompt that only instructs the LLM to 'Consider the strength of arguments when making recommendations.' No experiment, ablation, or formal argument shows that accepting, rejecting, modifying, or adding arguments changes the final priority list or intervention content. The stated limitation (scores 'support review rather than autonomous decision-making') does not address whether the scores affect the plan at all. This is load-bearing because the paper's headline contribution—preserving human authority—depends on human edits having a causal effect on the output.
[Section 3.2, Eq. (3)] The aggregation function g in Eq. (3) is never specified. The paper states that F(o_i) increases with strong support and decreases with strong challenges, but no concrete functional form, weights, or normalization is given. Likewise, the influence weights α_{y,x}, β_{y,x} and the squashing function σ in Eq. (2) are free parameters with no suggested values or sensitivity analysis. Without these specifications, the formal QBAF component is underdetermined, and the claim that F(o_i) is 'used by the system when constructing the final care plan' cannot be independently checked.
[Section 3.2, Argument Validation and Stage 3] The relationship between the intrinsic strength τ(x) defined before human review and the 'initial validity values' produced by the LLM-based Argument Validation prompt is unclear. Stage 3 says the system re-applies QBAF to ΓH using the revised argument set, and then the validation prompt evaluates 'each argument according to factual accuracy, relevance, practical feasibility, and evidence-based reasoning.' The text states the resulting scores 'are used as the initial validity values' before combining with relations, but it does not explain whether these replace τ(x) or are merged with it. This ambiguity makes the pipeline hard to reproduce and the claimed quantitative semantics difficult to verify.
[Section 4 (Results)] The 'Results' section contains only a single demonstrative scenario with no evaluation of the system's effectiveness. There is no user study, no comparison against a baseline (e.g., a non-contestable LLM planner), no measurement of whether the QBAF scores actually alter the final plan, and no assessment of whether human reviewers find the interface usable or trustworthy. Because the paper's claims concern preserving human agency and clinical accountability, a qualitative or quantitative evaluation of the contestation workflow is needed to substantiate those claims.
[Section 2.2 and Section 5] The LLM validity scores are self-referential in the sense that the same model family generates the arguments, evaluates their validity, and synthesizes the final plan. No external grounding, such as comparison with human expert ratings or clinical outcome data, is provided. Even if the QBAF machinery works correctly, unreliable τ(x) or F(o_i) values would be inherited by the final recommendations. The paper acknowledges in Section 5 that scores 'support review rather than autonomous decision-making,' but it does not address the risk that the scores are systematically biased, which is a correctness concern for any downstream use.

minor comments (5)

[Abstract/Keywords] The keyword 'graphic user interface' should be 'graphical user interface'.
[Stage 3, Section 3.2] The text states that human edits produce a revised argument set ΓH ⊆ Γ, but since reviewers can add new arguments, ΓH is not necessarily a subset of Γ. This should be clarified as ΓH = Γ' ∪ A, where A contains newly added arguments.
[Section 3.2, Eq. (2)] The parenthetical explanation 'all elements in each relation have a pair that contains the target x and the related y argument' is confusingly worded and should be rewritten for clarity.
[References] Several references appear to be incomplete (e.g., refs 5, 11, 16, 22 lack full author lists or page ranges). The authors should verify the citation details, especially for conference and journal papers.
[Section 4] The figure captions for Figs. 2 and 5 are repeated nearly verbatim; one should be shortened or made more informative with respect to the specific stage being shown.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: CoPlan is a reviewable pipeline, not a derivation whose output equals its input by construction.

full rationale

The paper makes no first-principles prediction or formal derivation; it describes a multi-stage decision-support interface. The quantitative quantities τ(x), f(x), and F(oi) are computed from the stated QBAF equations (Eqs. 2-3) and from LLM prompts, and the final care plan is produced by a separate LLM prompt. No equation in the paper equates the final plan to F(oi) or to τ(x), so there is no self-definitional reduction. The human-in-the-loop stage (Stage 3) edits Γ into ΓH and re-runs the semantics, providing an independent intervention point that breaks any closed loop. The self-citations in the related-work discussion (e.g., [19]) are contextual design claims supported by external references and do not carry a derivation, so they are not load-bearing. The skeptic's concern—that the final LLM prompt may not causally implement the updated scores—is a verification/evaluation gap, and the paper itself flags that validity and confidence scores support review rather than autonomous decision-making. A missing guarantee or external validation is a correctness risk, not circularity under the definitions in this analysis.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The central claim rests on standard argumentation semantics with unspecified influence weights and aggregation, on interRAI as a validated input, and on the assumption that LLM scoring is meaningful. No new physical or conceptual entities are introduced. The paper itself concedes in Section 5 that validity and confidence scores support review rather than autonomous decision-making.

free parameters (3)

Influence weights α_y,x and β_y,x = not specified
Equation (2) introduces non-negative weights controlling support and challenge influence, but no values or learning procedure are provided.
Aggregation function g = not specified
Equation (3) uses an unspecified aggregation function combining support and challenge degrees into option-level scores.
Squashing function σ = not specified
The iterative update f(x)=σ(τ(x)+I(x,f)) requires a squashing function that is not defined.

assumptions (3)

standard math QBAF semantics converge to a stable degree vector under iteration.
The system relies on existing gradual argumentation theory from Baroni et al. [5]; the convergence is assumed from that literature.
domain assumption interRAI Home Care assessment provides a valid and sufficient patient representation.
Stage 1 uses interRAI as the sole patient input, assuming it captures the clinical, functional, and psychosocial context needed for care planning.
ad hoc to paper LLM-generated validity scores reflect argument quality.
The 'Argument Validation prompt' asks an LLM to score factual accuracy and relevance, with no verification against expert judgments or outcome data.

how reviews work

0 comments

Cite this review

Pith. "Pith review of CoPlan: A Trustworthy Co-Intelligence Interface for Care Planning through Role-Based Contestable Argument Graphs." pith.science (2026). https://pith.science/paper/XBCNT6TS

@misc{pith2026260805107,
  author       = {Pith},
  title        = {Pith review of: CoPlan: A Trustworthy Co-Intelligence Interface for Care Planning through Role-Based Contestable Argument Graphs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XBCNT6TS}},
  note         = {Machine review of arXiv:2608.05107}
}

read the original abstract

AI-supported care planning can help clinicians, patients, caregivers, and care teams coordinate complex decisions across clinical, functional, psychosocial, and environmental needs. However, many AI systems present recommendations as fixed outputs, limiting stakeholders' ability to inspect, challenge, and revise plans when they conflict with clinical judgment, patient values, or real-world feasibility. We present CoPlan - a Co-Intelligent and Contestable Interface for Human-AI Care Planning. CoPlan uses a multi-agent workflow in which specialized AI agents generate candidate interventions and supporting or challenging arguments, while human care planners can accept, reject, modify, or add arguments before final plan generation. Through this design, CoPlan combines co-intelligence, in which humans and AI agents contribute complementary expertise, with contestability, where recommendations remain open to inspection, revision, and justification. We demonstrate CoPlan in an aging-in-place care planning scenario. The system supports adaptive care team recruitment, role-based argument review, final care plan generation, and practical follow-up through scheduling agents. This work contributes a contestable care planning interface and a design framing for trustworthy human-AI care planning that preserves human agency and clinical accountability.

Figures

Figures reproduced from arXiv: 2608.05107 by the authors.

**Figure 1.** Human-AI Co-Intelligence Mechanism through Contestability for Agentic AI in Care Planning Systems. The framework is guided by two principles: co-intelligence and contestability. Co-intelligence positions AI agents and human stakeholders as collaborative contributors with distinct knowledge and responsibilities. Contestability ensures that recommendations remain open to inspection, disagreement, revision, and justif… view at source ↗

**Figure 2.** Our proposed Role-based Contestable Argument Graph Computation Workflow for CoPlan Interface in Aging-in-Place Context. supporting and challenging arguments for each option, creating an argumentative pool Γ. A human reviewer may revise this set, producing ΓH. A validation operator V , grounded in quantitative bipolar argumentation semantics, assigns each argument a degree of acceptability. Finally, a care plan oper… view at source ↗

**Figure 3.** Interactive Graph Review Panel for role-based human-in-the-loop contestation. The interface allows reviewers to inspect support and challenge arguments, add new arguments, edit existing ones, remove inappropriate nodes, and confirm the revised argumentative structure before proceeding. Stage 3 – Role-Based Human-in-the-loop Contestation The goal of this stage is to introduce structured contestability while ensuring … view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Final Care Plan Viewer (right panel) and Chatbot Assistant (left panel) for further discussion. plan is a prescriptive, evidence-based set of recommendations tailored to the client’s physical, psychosocial, and environmental context. CoPlan uses the validated arguments…

**Figure 5.** Figure 5: Demonstrative result of CoPlan in an aging-in-place care planning scenario [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

28 extracted references · 25 canonical work pages

[1]

Minds and Machines33(4), 613–639 (2023), publisher: Springer

Alfrink, K., Keller, I., Kortuem, G., Doorn, N.: Contestable AI by design: Towards a framework. Minds and Machines33(4), 613–639 (2023), publisher: Springer

work page 2023
[2]

In: Proceedings of the 2019 chi conference on human factors in computing systems

Amershi, S., et al.: Guidelines for human-ai interaction. In: Proceedings of the 2019 chi conference on human factors in computing systems. pp. 1–13 (2019)

work page 2019
[3]

In: 12th International Conference on Autonomous Agents and Multiagent Systems 2013 (2013)

Amir, O., et al.: Collaborative health care plan support. In: 12th International Conference on Autonomous Agents and Multiagent Systems 2013 (2013)

work page 2013
[4]

anthropic.com/news/model-context-protocol

Anthropic: Introducing the model context protocol (2024),https://www. anthropic.com/news/model-context-protocol

work page 2024
[5]

International Journal of Approximate Rea- soning105(2019)

Baroni, P., et al.: From fine-grained properties to broad principles for gradual ar- gumentation: A principled spectrum. International Journal of Approximate Rea- soning105(2019)

work page 2019
[6]

Federal Practitioner42(5) (2025)

Borkowski, A.A., Ben-Ari, A.: Multiagent ai systems in health care: Envisioning next-generation intelligence. Federal Practitioner42(5) (2025)

work page 2025
[7]

arXiv preprint arXiv:2602.18916 (2026)

Cao, H.L., et al.: Adaptive collaboration of arena-based argumentative llms for ex- plainable and contestable legal reasoning. arXiv preprint arXiv:2602.18916 (2026)

arXiv 2026
[8]

In: Proc

Dignum, V., et al.: Contesting black-box ai decisions. In: Proc. of the 24th Inter- national Conference on Autonomous Agents and Multiagent Systems (2025)

work page 2025

Show all 28 references

[9]

Regulation (EU)679, 2016 (2016)

European Union: Regulation (eu) 2016/679 of the european parliament and of the council. Regulation (EU)679, 2016 (2016)

2016
[10]

The EU Artificial Intelligence Act (2024)

European Union: The eu artificial intelligence act. The EU Artificial Intelligence Act (2024)

2024
[11]

In: Proceedings of the AAAI Conference on Arti- ficial Intelligence

Freedman, G., et al.: Argumentative Large Language Models for Explainable and Contestable Claim Verification. In: Proceedings of the AAAI Conference on Arti- ficial Intelligence. vol. 39, pp. 14930–14939 (2025), issue: 14

2025
[12]

BMC health services research9(1), 71 (2009)

Gray, L.C., et al.: Sharing clinical information across care settings: the birth of an integrated assessment system. BMC health services research9(1), 71 (2009)

2009
[13]

Science robotics4(37), eaay7120 (2019)

Gunning, D., Stefik, M., Choi, J., Miller, T., Stumpf, S., Yang, G.Z.: XAI—Explainable artificial intelligence. Science robotics4(37), eaay7120 (2019)

2019
[14]

In: 2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM)

Hong, S., et al.: Argmed-agents: explainable clinical decision reasoning with llm disscusion via argumentation schemes. In: 2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM). pp. 5486–5493. IEEE (2024)

2024
[15]

Advances in Neural Information Processing Systems37, 79410–79452 (2024)

Kim, Y., et al.: Mdagents: An adaptive collaboration of llms for medical decision- making. Advances in Neural Information Processing Systems37, 79410–79452 (2024)

2024
[16]

In: Pro- ceedings of the 21st International Conference on Principles of Knowledge Repre- sentation and Reasoning

Leofante, F., et al.: Contestable ai needs computational argumentation. In: Pro- ceedings of the 21st International Conference on Principles of Knowledge Repre- sentation and Reasoning. pp. 888–896 (2024)

2024
[17]

arXiv preprint arXiv:2405.02957 (2024)

Li, J., et al.: Agent hospital: A simulacrum of hospital with evolvable medical agents. arXiv preprint arXiv:2405.02957 (2024)

2024 arXiv
[18]

In: Proceedings of the 1st FPT AI Conference, FAIC 2021

Nguyen, H., Cao, H.Q., Nguyen, K.V.T., Pham, N.D.K.: Evaluation of explainable artificial intelligence: Shap, lime, and cam. In: Proceedings of the 1st FPT AI Conference, FAIC 2021. pp. 1–6. Hanoi, Vietnam (2021)

2021
[19]

ACM Trans

Nguyen, H., Rahimi, A., Whitford, V., Fournier, H., Kondratova, I., Richard, R., Cao, H.: Heart2mind: Human-centered contestable psychiatric disorder prediction system using wearable ecg monitors. ACM Trans. Comput. Healthcare (2026)

2026
[20]

In: Proceedings of 9th International Symposium on Chatbots and Human-centred AI 2025 (2025)

Nguyen, L.P.T., et al.: Motion2Meaning: A Clinician-Centered Framework for Con- testable LLM in Parkinson’s Disease Gait Interpretation. In: Proceedings of 9th International Symposium on Chatbots and Human-centred AI 2025 (2025)

2025
[21]

WHO guidance

Organization, W.H.: Ethics and governance of artificial intelligence for health: large multi-modal models. WHO guidance. World Health Organization (2024)

2024
[22]

Artificial intelligence in medicine107(2020)

Ploug, T., Holm, S.: The four dimensions of contestable AI diagnostics-A patient- centric approach to explainable AI. Artificial intelligence in medicine107(2020)

2020
[23]

ACM Transactions on Computer-Human Interaction30(2), 1–34 (2023)

Procter, R., et al.: Holding ai to account: challenges for the delivery of trustworthy ai in healthcare. ACM Transactions on Computer-Human Interaction30(2), 1–34 (2023)

2023
[24]

Human behavior and emerging technologies1(1), 33–36 (2019)

Riedl, M.O.: Human-centered artificial intelligence and machine learning. Human behavior and emerging technologies1(1), 33–36 (2019)

2019
[25]

International Journal of Human–Computer Interaction36(6), 495–504 (2020)

Shneiderman, B.: Human-centered artificial intelligence: Reliable, safe & trustwor- thy. International Journal of Human–Computer Interaction36(6), 495–504 (2020)

2020
[26]

Canadian Family Physi- cian61(5), 449–453 (2015)

Stewart, M., Ryan, B.: Ecology of health care in canada. Canadian Family Physi- cian61(5), 449–453 (2015)

2015
[27]

In: Findings of the Association for Computational Linguistics: ACL 2024 (2024)

Tang, X., et al.: Medagents: Large language models as collaborators for zero-shot medical reasoning. In: Findings of the Association for Computational Linguistics: ACL 2024 (2024)

2024
[28]

Nature medicine25(1), 44–56 (2019)

Topol, E.J.: High-performance medicine: the convergence of human and artificial intelligence. Nature medicine25(1), 44–56 (2019)

2019

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

Minds and Machines33(4), 613–639 (2023), publisher: Springer

Alfrink, K., Keller, I., Kortuem, G., Doorn, N.: Contestable AI by design: Towards a framework. Minds and Machines33(4), 613–639 (2023), publisher: Springer

work page 2023

[2] [2]

In: Proceedings of the 2019 chi conference on human factors in computing systems

Amershi, S., et al.: Guidelines for human-ai interaction. In: Proceedings of the 2019 chi conference on human factors in computing systems. pp. 1–13 (2019)

work page 2019

[3] [3]

In: 12th International Conference on Autonomous Agents and Multiagent Systems 2013 (2013)

Amir, O., et al.: Collaborative health care plan support. In: 12th International Conference on Autonomous Agents and Multiagent Systems 2013 (2013)

work page 2013

[4] [4]

anthropic.com/news/model-context-protocol

Anthropic: Introducing the model context protocol (2024),https://www. anthropic.com/news/model-context-protocol

work page 2024

[5] [5]

International Journal of Approximate Rea- soning105(2019)

Baroni, P., et al.: From fine-grained properties to broad principles for gradual ar- gumentation: A principled spectrum. International Journal of Approximate Rea- soning105(2019)

work page 2019

[6] [6]

Federal Practitioner42(5) (2025)

Borkowski, A.A., Ben-Ari, A.: Multiagent ai systems in health care: Envisioning next-generation intelligence. Federal Practitioner42(5) (2025)

work page 2025

[7] [7]

arXiv preprint arXiv:2602.18916 (2026)

Cao, H.L., et al.: Adaptive collaboration of arena-based argumentative llms for ex- plainable and contestable legal reasoning. arXiv preprint arXiv:2602.18916 (2026)

arXiv 2026

[8] [8]

In: Proc

Dignum, V., et al.: Contesting black-box ai decisions. In: Proc. of the 24th Inter- national Conference on Autonomous Agents and Multiagent Systems (2025)

work page 2025