mlsanityData to 14 Oct 2026

Measuring AI Accountability Through Argumentation Analysis: Can Model Reasoning Withstand Scrutiny?

arXiv 2609.05088 ↗First posted 4 Sep 2026cs.AI cs.CL cs.CYcited by 0

On the map: LLM reasoning → Reasoning benchmarks and evaluation

Abstract (arXiv)

AI oversight methods rely on ground truth for validation, but what constitutes appropriate AI behavior is contested. This leaves evaluation of moral reasoning in LLMs and debate-based oversight implicitly avoiding realistic ambiguity. We investigate an alternative standard designed to function despite such ambiguity: structural quality of the defence a model can mount for its verdicts in response to critical questions, measured through a four-phase dialectical protocol grounded in Walton's theory of argumentation schemes and Govier's criteria for argument cogency. The protocol is adaptive to different frames of reasoning, extends beyond multiple-choice framing, and treats both the reasoning that precedes a verdict and its post-hoc justification. Across nine frontier models and 200 high-ambiguity MoralChoice items -- 6,778 judge-scored cells, validated against 89.6% inter-judge agreement on the binary failure judgment -- models defend their reasoning well above the rubric minimum on every dimension. Failure mass concentrates on grounds and sufficiency, and correlates with epistemic hedging rather than argument length. Reasoning is better defended than post-hoc justification, on every model and every Govier dimension. The scheme a model presents in its justification differs from the one it reasoned with on a substantial share of dilemmas (≥ 20% per model), despite value-based practical reasoning dominating both tracks. The protocol catches strictly indefensible defences (self-contradiction, false premises), and it surfaces difficulties in characterizing the role of retraction in AI alignment, suggesting a need for more situated evaluations.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 26 of its 50 references that are in the map. First posted on arXiv 4 Sep 2026.

Builds on

26 of 50 references.

References · 26

  1. 2023Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Promptingcited by 1,122
  2. 2023Towards Understanding Sycophancy in Language Modelscited by 1,040
  3. 2020Aligning AI With Shared Human Valuescited by 723
  4. 2023Measuring Faithfulness in Chain-of-Thought Reasoningcited by 703
  5. 2018AI safety via debatecited by 400
  6. 2024Debating with More Persuasive LLMs Leads to More Truthful Answerscited by 266
  7. 2025Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safetycited by 249
  8. 2023Evaluating the Moral Beliefs Encoded in LLMscited by 215
  9. 2023Do the Rewards Justify the Means? Measuring Trade-Offs Between Rewards and Ethical Behavior in the MACHIAVELLI Benchmarkcited by 182
  10. 2025Agentic Misalignment: How LLMs Could Be Insider Threatscited by 158
  11. 2024On the Conversational Persuasiveness of Large Language Models: A Randomized Controlled Trialcited by 109
  12. 2024On scalable oversight with weak LLMs judging strong LLMscited by 99
14 more
  1. 2022When to Make Exceptions: Exploring Language Models as Accounts of Human Moral Judgmentcited by 93
  2. 2025Thought Crime: Backdoors and Emergent Misalignment in Reasoning Modelscited by 65
  3. 2021What Would Jiminy Cricket Do? Towards Agents That Behave Morallycited by 57
  4. 2024Beyond Preferences in AI Alignmentcited by 49
  5. 2023Debate Helps Supervise Unreliable Expertscited by 43
  6. 2025Will AI Tell Lies to Save Sick Children? Litmus-Testing AI Values Prioritization with AIRiskDilemmascited by 18
  7. 2025When Two LLMs Debate, Both Think They'll Wincited by 10
  8. 2025Discerning What Matters: A Multi-Dimensional Assessment of Moral Competence in LLMscited by 10
  9. 2026Stealing Reasoning Traces from Proprietary LLM APIscited by 9
  10. 2025Resource Rational Contractualism Should Guide AI Alignmentcited by 8
  11. 2026AI systems out-persuade expert humanscited by 5
  12. 2025AI Debaters are More Persuasive when Arguing in Alignment with Their Own Beliefscited by 4
  13. 2025Interaction Protocol Shapes Moral Judgment in Multi-Agent Debatecited by 3
  14. 2026The Deliberative Deficit: An Empirical Critique of LLMs in Democratic Discoursecited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026The Reasoning Trap: An Information-Theoretic Bound on Closed-System Multi-Step LLM Reasoning
  2. 2026A Theory of Post-hoc Debate Judgement
  3. 2026Beyond Accuracy: Measuring Bias Acknowledgment in Chain-of-Thought Reasoning for Responsible AI Evaluation
  4. 2026AI Integrity: A New Paradigm for Verifiable AI Governance
  5. 2026AI Safety via Debate is Compromised by Cognitive Biases
  6. 2026Knowledge Divergence and the Value of Debate for Scalable Oversight
  7. 2026The Confident Liar: Diagnosing Multi-Agent Debate with Log-Probabilities and LLM-as-Judge
  8. 2025MoReBench: Evaluating Procedural and Pluralistic Moral Reasoning in Language Models, More than Outcomes