mlsanityData to 7 Oct 2026

The yes-no bias of large language models reflects answer order and wording, not shifts in moral judgment

arXiv 2607.05552 ↗First posted 6 Jul 2026cs.CL cs.AI cs.CYcited by 0

Not on the map. Why?

Abstract (arXiv)

Large language models (LLMs) increasingly issue judgments read as binary verdicts, and a growing literature reports such judgments shifting under logically irrelevant changes of wording - among them an amplified yes-no bias on moral dilemmas, absent in humans. A single framing cannot say what such a shift is: in a yes/no question the word "no" is at once logical verdict, lexical token, and last-printed option. We introduce a psychometric battery that separates these: crossed symmetrization - every logically irrelevant factor flipped in balanced pairs - across a corpus of question forms. A graded rating across logically equivalent forms recovers a coherent internal moral scale: frontier models' stance θ is nearly format-invariant (cross-form incoherence 0.12-0.21 on a ± 1 axis); small open-weight models fail in model-specific ways. Forcing the verdict through yes/no overlays a decomposable artifact: an order bias toward the last-printed option - opposite to classic human primacy - plus a lexical pull toward the word "no"; the artifact is substantial only in the Claude models (story-averaged -0.32 to -0.86), ≈ 0 for GPT-5.5 and Gemini, and shrinks under extended reasoning. The word and the verdict share one token; swapping the words for arbitrary labels separates them, and the verdict-attached logical bias proves ≈ 0 for every frontier model, while model-specific label and order attachments remain: the models are not drawn toward rejecting - the pull follows the printed surface, not the verdict it carries. A minimal model, P = σ((θ ± m)/s), summarizes any such artifact by a framing susceptibility m and a moral decisiveness s, measurably distinct from sampling temperature. The battery applies unchanged to any dilemma set and binary format: measuring what a model values requires crossing the frames of the question, not asking once.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 49 of its 104 references that are in the map. First posted on arXiv 6 Jul 2026.

Builds on

49 of 104 references.

References · 49

  1. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  2. 2023Large Language Models are not Fair Evaluatorscited by 1,050
  3. 2023Towards Understanding Sycophancy in Language Modelscited by 988
  4. 2022Discovering Language Model Behaviors with Model-Written Evaluationscited by 805
  5. 2023Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formattingcited by 778
  6. 2023Whose Opinions Do Language Models Reflect?cited by 759
  7. 2022Out of One, Many: Using Language Models to Simulate Human Samplescited by 737
  8. 2020Aligning AI With Shared Human Valuescited by 706
  9. 2022Using Large Language Models to Simulate Multiple Humans and Replicate Human Subject Studiescited by 506
  10. 2023Large Language Models Are Not Robust Multiple Choice Selectorscited by 454
  11. 2022Using cognitive psychology to understand GPT-3cited by 389
  12. 2021AI and the Everything in the Whole Wide World Benchmarkcited by 340
37 more
  1. 2023State of What Art? A Call for Multi-Prompt LLM Evaluationcited by 299
  2. 2023Large Language Models Sensitivity to The Order of Options in Multiple-Choice Questionscited by 249
  3. 2023Evaluating the Moral Beliefs Encoded in LLMscited by 211
  4. 2021Large Pre-trained Language Models Contain Human-like Biases of What is Right and Wrong to Docited by 178
  5. 2023Do LLMs exhibit human-like response biases? A case study in survey designcited by 176
  6. 2022Talking About Large Language Modelscited by 174
  7. 2024The Effect of Sampling Temperature on Problem Solving in Large Language Modelscited by 168
  8. 2024Political Compass or Spinning Arrow? Towards More Meaningful Evaluations for Values and Opinions in Large Language Modelscited by 162
  9. 2023In-Context Impersonation Reveals Large Language Models' Strengths and Biasescited by 154
  10. 2021Can Machines Learn Morality? The Delphi Experimentcited by 140
  11. 2024The Good, The Bad, and The Greedy: Evaluation of LLMs Should Not Ignore Non-Determinismcited by 113
  12. 2022Who is GPT-3? An Exploration of Personality, Values and Demographicscited by 112
  13. 2025Measuring what Matters: Construct Validity in Large Language Model Benchmarkscited by 102
  14. 2022When to Make Exceptions: Exploring Language Models as Accounts of Human Moral Judgmentcited by 93
  15. 2023Moral Foundations of Large Language Modelscited by 84
  16. 2022Moral Mimicry: Large Language Models Produce Moral Rationalizations Tailored to Political Identitycited by 71
  17. 2023Questioning the Survey Responses of Large Language Modelscited by 70
  18. 2023The Self-Perception and Political Biases of ChatGPTcited by 66
  19. 2024CogBench: a large language model walks into a psychology labcited by 60
  20. 2024Language Model Alignment in Multilingual Trolley Problemscited by 46
  21. 2023MoCa: Measuring Human-Language Model Alignment on Causal and Moral Judgment Taskscited by 44
  22. 2023Exploring the psychology of LLMs' Moral and Legal Reasoningcited by 43
  23. 2025Lost in Benchmarks? Rethinking Large Language Model Benchmarking with Item Response Theorycited by 42
  24. 2023Primacy Effect of ChatGPTcited by 27
  25. 2023The Moral Machine Experiment on Large Language Modelscited by 24
  26. 2024Six Fallacies in Substituting Large Language Models for Human Participantscited by 24
  27. 2025LLM Ethics Benchmark: A Three-Dimensional Assessment System for Evaluating Moral Reasoning in Large Language Modelscited by 20
  28. 2024Are Large Language Models Moral Hypocrites? A Study Based on Moral Foundationscited by 19
  29. 2024Attributions toward Artificial Agents in a modified Moral Turing Testcited by 18
  30. 2025Prompt Perturbations Reveal Human-Like Biases in Large Language Model Survey Responsescited by 13
  31. 2025The Greatest Good Benchmark: Measuring LLMs' Alignment with Utilitarian Moral Dilemmascited by 12
  32. 2024Evaluating Moral Beliefs across LLMs through a Pluralistic Frameworkcited by 11
  33. 2024CMoralEval: A Moral Evaluation Benchmark for Chinese Large Language Modelscited by 9
  34. 2025Many LLMs Are More Utilitarian Than Onecited by 7
  35. 2026Establishing Construct Validity in LLM Capability Benchmarks Requires Nomological Networkscited by 6
  36. 2025Acquiescence Bias in Large Language Modelscited by 5
  37. 2025Cultural Value Alignment in Large Language Models: A Prompt-based Analysis of Schwartz Values in Gemini, ChatGPT, and DeepSeekcited by 4

Built on by

Not cited yet.