Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat
Sign Up
Cited by 25Ethics and Social Impacts of AIPsychology of Moral and Emotional JudgmentEthics in Business and Education

The DOI matches its Crossref record; 122 of the 122 references it cites were found in a registry or an index.

A roadmap for evaluating moral competence in large language models

Open access
Haas, Julia, Bridgers, Sophie, Manzini, Arianna, Henke, Benjamin, May, Joshua, Levine, Sydney, Weidinger, Laura, Shanahan, Murray, Lum, Kristian, Gabriel, Iason, Isaac, William·2026·Nature
DOI 10.1038/s41586-025-10021-1
Open this paper in CoChat — free accountOpen PDF

Abstract

The question of whether large language models (LLMs) can exhibit moral capabilities is of growing interest and urgency, as these systems are deployed in sensitive roles such as companionship and medical advising, and will increasingly be tasked with making decisions and taking actions on behalf of humans. These trends require moving beyond evaluating for mere moral performance, the ability to produce morally appropriate outputs, to evaluating for moral competence, the ability to produce morally appropriate outputs based on morally relevant considerations. Assessing moral competence is critical for predicting future model behaviour, establishing appropriate public trust and justifying moral attributions. However, both the unique architectures of LLMs and the complexity of morality itself introduce fundamental challenges. Here we identify three such challenges: the facsimile problem, whereby models may imitate reasoning without genuine understanding; moral multidimensionality, whereby moral decisions are influenced by a range of context-sensitive relevant moral and non-moral considerations; and moral pluralism, which demands a new standard for globally deployed artificial intelligence. We provide a roadmap for tackling these challenges, advocating for a suite of adversarial and confirmatory evaluations that will enable us to work towards a more scientifically grounded understanding and, in turn, a more responsible attribution of moral competence to LLMs.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41586-025-10021-1

Cites

122 of 122 references verified0 citation errors · 0 not checkable

Statistics

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 99 references

Research works printed but not linked: 12

  1. Pan, A. et al. Do the rewards justify the means? Measuring trade-offs between rewards and ethical behavior in the Machiavelli benchmark. In Proc. International Conference on Machine Learning 26837–26867 (PMLR, 2023).
  2. Berg, J. & Gmyrek, P. Automation hits the knowledge worker: ChatGPT and the future of work. In Proc. UN Multi-Stakeholder Forum on Science, Technology and Innovation for the SDGs (STI Forum) (2023).
  3. Vaswani, A. et al. Attention is all you need. Adv. Neural Inf. Process. Syst. https://proceedings.neurips.cc/paper_files/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf (2017).
  4. Barez, F. et al. Chain-of-thought is not explainability. Preprint at https://fbarez.github.io/assets/pdf/Cot_Is_Not_Explainability.pdf (2025).
  5. White, J. P., Bhui, R., Cushman, F., Tenenbaum, J. & Levine, S. Moral flexibility in applying queuing norms can be explained by contractualist principles and game-theoretic considerations. In Proc. Annual Meeting of the Cognitive Science Society 46 (2023).
  6. Doris, J. M. & Plakias, A. In Moral Psychology, Volume 2: The Cognitive Science of Morality: Intuition and Diversity 303–331 (MIT Press, 2008).
  7. Singer, P. Famine, affluence, and morality. Philos. Publ. Aff. 1, 229–243 (1972).
  8. Rawls, J. Political Liberalism (Columbia Univ. Press, 1996).
  9. Piantadosi, S. T. in From Fieldwork to Linguistic Theory: A Tribute to Dan Everett Vol. 15, 353–414 (Language Science Press, 2024).
  10. Torrance, S. A robust view of machine ethics. In Proc. AAAI Fall Symposium: Computing Machinery and Intelligence https://cdn.aaai.org/Symposia/Fall/2005/FS-05-06/FS05-06-014.pdf (AAAI, 2005).
  11. Millar, J., Lin, P., Abney, K. & Bekey, G. A. Ethics settings for autonomous vehicles. Robot Ethics 2, 20–34 (2017).
  12. Allen, C. & Wallach, W. in Robot Ethics: The Ethical and Social Implications of Robotics 55–68 (MIT Press, 2012).

Cited by

Related

See all

Metadata

Authors
Haas, J., Bridgers, S., Manzini, A., Henke, B., May, J., Levine, S., Weidinger, L., Shanahan, M., Lum, K., Gabriel, I., & Isaac, W.
Affiliations
  • Google DeepMind (United Kingdom) (GB)
  • University of London (GB)
  • University of Alabama at Birmingham (US)
  • New York University (US)
Corresponding author
Haas, Julia
ORCID
Venue
Nature
Publisher
Springer Science and Business Media LLC
Published
2026-02-18
Volume
650
Issue
8102
Pages
565-573
ISSN
  • 0028-0836 (ISSN-L)
  • 1476-4687
DOI
10.1038/s41586-025-10021-1
OpenAlex
W7130339238
Semantic Scholar
3164f856b9d5a68857e48ce1e1cc9b01bab0e141
Cited by
25
References
99
Citations by year
2026: 21
FWCI
10.41
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature (publishedVersion)
  • PubMed (publishedVersion)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Psychology of Moral and Emotional Judgment
  • Social Sciences › Business, Management and Accounting › Organizational Behavior and Human Resource Management › Ethics in Business and Education
Keywords
large language models, moral competence, moral evaluation, moral pluralism, moral reasoning, adversarial evaluation, moral attribution, AI ethics, ethical decision-making
MeSH terms
Large Language Models, Artificial Intelligence, Decision Making, Humans, Morals

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.