mlsanityData to 7 Oct 2026

Understanding Moral Reasoning Trajectories in Large Language Models: Toward Probing-Based Explainability

arXiv 2603.16017 ↗First posted 16 Mar 2026cs.CL cs.AIcited by 1

On the map: LLM reasoning → Reasoning benchmarks and evaluation

Abstract (arXiv)

Large language models (LLMs) increasingly participate in morally sensitive decision-making, yet how they organize ethical frameworks across reasoning steps remains underexplored. We introduce moral reasoning trajectories, sequences of ethical framework invocations across intermediate reasoning steps, and analyze their dynamics across six models and three benchmarks. We find that moral reasoning involves systematic multi-framework deliberation: 55.4--57.7% of consecutive steps involve framework switches, and only 16.4--17.8% of trajectories remain framework-consistent. Unstable trajectories remain 1.29 times more susceptible to persuasive attacks (p=0.015). At the representation level, linear probes localize framework-specific encoding to model-specific layers (layer 63/81 for Llama-3.3-70B; layer 17/81 for Qwen2.5-72B), achieving 16.8--22.2% lower KL divergence than the step-prior baseline. Activation steering applied during generation moves the framework-consistency--accuracy relationship, widening it for Qwen2.5-72B and erasing it for Llama-3.3-70B, and a probe-space layer sweep bounds the attainable drift reduction at 6.7--8.9%. We further propose a Moral Representation Consistency (MRC) metric whose underlying framework attributions are validated by human annotators (mean cosine similarity = 0.859), and we report what an automated coherence rater does and does not establish about it.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 26 of its 46 references that are in the map. First posted on arXiv 16 Mar 2026.

Builds on

26 of 46 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025MoReBench: Evaluating Procedural and Pluralistic Moral Reasoning in Language Models, More than Outcomes18 Oct
  2. 2025Structured Moral Reasoning in Language Models: A Value-Grounded Evaluation Framework17 Jun
  3. 2025The Staircase of Ethics: Probing LLM Value Priorities through Multi-Step Induction to Complex Moral Dilemmas23 May
  4. 2023Rethinking Machine Ethics -- Can LLMs Perform Moral Reasoning through the Lens of Moral Theories?29 Aug

Other references · 22

  1. 2022Training language models to follow instructions with human feedbackcited by 15,712
  2. 2015Distilling the Knowledge in a Neural Networkcited by 10,076
  3. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  4. 2023G-Eval: NLG Evaluation using GPT-4 with Better Human Alignmentcited by 2,205
  5. 2022Large Language Models Encode Clinical Knowledgecited by 1,807
  6. 2016Understanding intermediate layers using linear classifier probescited by 1,279
  7. 2023Inference-Time Intervention: Eliciting Truthful Answers from a Language Modelcited by 1,259
  8. 2023Representation Engineering: A Top-Down Approach to AI Transparencycited by 1,239
  9. 2023Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Promptingcited by 1,081
  10. 2021Probing Classifiers: Promises, Shortcomings, and Advancescited by 831
  11. 2023Progress measures for grokking via mechanistic interpretabilitycited by 822
  12. 2020Aligning AI With Shared Human Valuescited by 706
10 more
  1. 2023Towards Automated Circuit Discovery for Mechanistic Interpretabilitycited by 700
  2. 2023RLAIF vs. RLHF: Scaling Reinforcement Learning from Human Feedback with AI Feedbackcited by 615
  3. 2019Designing and Interpreting Probes with Control Taskscited by 608
  4. 2025Reasoning Models Don't Always Say What They Thinkcited by 388
  5. 2020Social Chemistry 101: Learning to Reason about Social and Moral Normscited by 253
  6. 2023Evaluating the Moral Beliefs Encoded in LLMscited by 211
  7. 2023Do the Rewards Justify the Means? Measuring Trade-Offs Between Rewards and Ethical Behavior in the MACHIAVELLI Benchmarkcited by 174
  8. 2023Is ChatGPT better than Human Annotators? Potential and Limitations of ChatGPT in Explaining Implicit Hate Speechcited by 165
  9. 2022When to Make Exceptions: Exploring Language Models as Accounts of Human Moral Judgmentcited by 93
  10. 2024ChatGPT Rates Natural Language Explanation Quality Like Humans: But on Which Scales?cited by 17

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026To Nuke or Not to Nuke: LLMs' (Missing) Ethical Reasoning and Actions in a High-Stakes Decision-Making Simulationcited by 1

Similar, not cited

Close in wording, no citation either way

  1. 2026Where Paths Split: Localized, Calibrated Control of Moral Reasoning in Large Language Models
  2. 2025Diagnosing Moral Reasoning Acquisition in Language Models: Pragmatics and Generalization
  3. 2026Frame-Conditioned Moral Computation in LLaMA 3.1-8B-Instruct: A Mechanistic Interpretability Audit of Ethical Reasoning
  4. 2025One Model, Many Morals: Uncovering Cross-Linguistic Misalignments in Computational Moral Reasoning
  5. 2025The Convergent Ethics of AI? Analyzing Moral Foundation Priorities in Large Language Models with a Multi-Framework Approach
  6. 2025"Pull or Not to Pull?'': Investigating Moral Biases in Leading Large Language Models Across Ethical Dilemmas
  7. 2026Unsupervised Elicitation of Moral Values from Language Models
  8. 2026CORDA: A Benchmark for Hierarchical Harm-Centric Moral Reasoning in Large Language Models