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

arXiv: https://arxiv.org/abs/2609.05088 · HTML: https://mlsanity.com/paper/2609.05088/

## 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.

## Placed in

- LLM reasoning / [Reasoning benchmarks and evaluation](https://mlsanity.com/lr/reasoning-benchmarks-and-evaluation/)

## Builds on (most cited first, 26 of 26)

- [Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting](https://mlsanity.com/paper/2305.04388/) (2023)
- [Towards Understanding Sycophancy in Language Models](https://mlsanity.com/paper/2310.13548/) (2023)
- [Aligning AI With Shared Human Values](https://mlsanity.com/paper/2008.02275/) (2020)
- [Measuring Faithfulness in Chain-of-Thought Reasoning](https://mlsanity.com/paper/2307.13702/) (2023)
- [AI safety via debate](https://mlsanity.com/paper/1805.00899/) (2018)
- [Debating with More Persuasive LLMs Leads to More Truthful Answers](https://mlsanity.com/paper/2402.06782/) (2024)
- [Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety](https://mlsanity.com/paper/2507.11473/) (2025)
- [Evaluating the Moral Beliefs Encoded in LLMs](https://mlsanity.com/paper/2307.14324/) (2023)
- [Do the Rewards Justify the Means? Measuring Trade-Offs Between Rewards and Ethical Behavior in the MACHIAVELLI Benchmark](https://mlsanity.com/paper/2304.03279/) (2023)
- [Agentic Misalignment: How LLMs Could Be Insider Threats](https://mlsanity.com/paper/2510.05179/) (2025)
- [On the Conversational Persuasiveness of Large Language Models: A Randomized Controlled Trial](https://mlsanity.com/paper/2403.14380/) (2024)
- [On scalable oversight with weak LLMs judging strong LLMs](https://mlsanity.com/paper/2407.04622/) (2024)
- [When to Make Exceptions: Exploring Language Models as Accounts of Human Moral Judgment](https://mlsanity.com/paper/2210.01478/) (2022)
- [Thought Crime: Backdoors and Emergent Misalignment in Reasoning Models](https://mlsanity.com/paper/2506.13206/) (2025)
- [What Would Jiminy Cricket Do? Towards Agents That Behave Morally](https://mlsanity.com/paper/2110.13136/) (2021)
- [Beyond Preferences in AI Alignment](https://mlsanity.com/paper/2408.16984/) (2024)
- [Debate Helps Supervise Unreliable Experts](https://mlsanity.com/paper/2311.08702/) (2023)
- [Will AI Tell Lies to Save Sick Children? Litmus-Testing AI Values Prioritization with AIRiskDilemmas](https://mlsanity.com/paper/2505.14633/) (2025)
- [When Two LLMs Debate, Both Think They'll Win](https://mlsanity.com/paper/2505.19184/) (2025)
- [Discerning What Matters: A Multi-Dimensional Assessment of Moral Competence in LLMs](https://mlsanity.com/paper/2506.13082/) (2025)
- [Stealing Reasoning Traces from Proprietary LLM APIs](https://mlsanity.com/paper/2608.09867/) (2026)
- [Resource Rational Contractualism Should Guide AI Alignment](https://mlsanity.com/paper/2506.17434/) (2025)
- [AI systems out-persuade expert humans](https://mlsanity.com/paper/2606.16475/) (2026)
- [AI Debaters are More Persuasive when Arguing in Alignment with Their Own Beliefs](https://mlsanity.com/paper/2510.13912/) (2025)
- [Interaction Protocol Shapes Moral Judgment in Multi-Agent Debate](https://mlsanity.com/paper/2510.10002/) (2025)
- [The Deliberative Deficit: An Empirical Critique of LLMs in Democratic Discourse](https://mlsanity.com/paper/2608.10186/) (2026)

## 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.