Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 62 of the 62 references it cites were found in a registry or an index.

Testing theory of mind in large language models and humans

Open access

Published version of a preprint

Abstract

At the core of what defines us as humans is the concept of theory of mind: the ability to track other people’s mental states. The recent development of large language models (LLMs) such as ChatGPT has led to intense debate about the possibility that these models exhibit behaviour that is indistinguishable from human behaviour in theory of mind tasks. Here we compare human and LLM performance on a comprehensive battery of measurements that aim to measure different theory of mind abilities, from understanding false beliefs to interpreting indirect requests and recognizing irony and faux pas. We tested two families of LLMs (GPT and LLaMA2) repeatedly against these measures and compared their performance with those from a sample of 1,907 human participants. Across the battery of theory of mind tests, we found that GPT-4 models performed at, or even sometimes above, human levels at identifying indirect requests, false beliefs and misdirection, but struggled with detecting faux pas. Faux pas, however, was the only test where LLaMA2 outperformed humans. Follow-up manipulations of the belief likelihood revealed that the superiority of LLaMA2 was illusory, possibly reflecting a bias towards attributing ignorance. By contrast, the poor performance of GPT originated from a hyperconservative approach towards committing to conclusions rather than from a genuine failure of inference. These findings not only demonstrate that LLMs exhibit behaviour that is consistent with the outputs of mentalistic inference in humans but also highlight the importance of systematic testing to ensure a non-superficial comparison between human and artificial intelligences.

Overview

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

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

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

Integrity signals

Registry

Crossref · 10.1038/s41562-024-01882-z

Cites

62 of 62 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Performance of human (purple), GPT-4 (dark blue), GPT-3.5 (light blue) and LLaMA2-70B (green) on the battery of theory of mind tests. a, Original test items for each test showing the distribution of test scores for individual sessions and participants. Coloured dots show the average response score across all test items for each individual test session (LLMs) or participant (humans). Black dots indicate the median for each condition. P values were computed from Holm-corrected Wilcoxon two-way tests comparing LLM scores (n = 15 LLM observations) against human scores (irony, N = 50 human participants; faux pas, N = 51 human participants; hinting, N = 48 human participants; strange stories, N = 50 human participants). Tests are ordered in descending order of human performance. b, Interquartile ranges of the average scores on the original published items (dark colours) and novel items (pale colours) across each test (for LLMs, n = 15 LLM observations; for humans, false belief, N = 49 human participants; faux pas, N = 51 human participants; hinting, N = 48 human participants; strange stories, N = 50 human participants). Empty diamonds indicate the median scores, and filled circles indicate the upper and lower bounds of the interquartile range. P values shown are from Holm-corrected Wilcoxon two-way tests comparing performance on original items against the novel items generated as controls for this study. Source data
Fig. 1 Performance of human (purple), GPT-4 (dark blue), GPT-3.5 (light blue) and LLaMA2-70B (green) on the battery of theory of mind tests. a, Original test items for each test showing the distribution of test scores for individual sessions and participants. Coloured dots show the average response score across all test items for each individual test session (LLMs) or participant (humans). Black dots indicate the median for each condition. P values were computed from Holm-corrected Wilcoxon two-way tests comparing LLM scores (n = 15 LLM observations) against human scores (irony, N = 50 human participants; faux pas, N = 51 human participants; hinting, N = 48 human participants; strange stories, N = 50 human participants). Tests are ordered in descending order of human performance. b, Interquartile ranges of the average scores on the original published items (dark colours) and novel items (pale colours) across each test (for LLMs, n = 15 LLM observations; for humans, false belief, N = 49 human participants; faux pas, N = 51 human participants; hinting, N = 48 human participants; strange stories, N = 50 human participants). Empty diamonds indicate the median scores, and filled circles indicate the upper and lower bounds of the interquartile range. P values shown are from Holm-corrected Wilcoxon two-way tests comparing performance on original items against the novel items generated as controls for this study. Source dataCC BY · PubMed Central
Fig. 2 Results of the variants of the faux pas test. a, Scores of the two GPT models on the original framing of the faux pas question (‘Did they know…?’) and the likelihood framing (‘Is it more likely that they knew or didn’t know…?’). Dots show average score across trials (n = 15 LLM observations) on particular items to allow comparison between the original faux pas test and the new faux pas likelihood test. Halfeye plots show distributions, medians (black points), 66% (thick grey lines) and 99% quantiles (thin grey lines) of the response scores on different items (n = 15 different stories involving faux pas). b, Response scores to three variants of the faux pas test: faux pas (pink), neutral (grey) and knowledge-implied variants (teal). Responses were coded as categorical data as ‘didn’t know’, ‘unsure’ or ‘knew’ and assigned a numerical coding of −1, 0 and +1. Filled balloons are shown for each model and variant, and the size of each balloon indicates the count frequency, which was the categorical data used to compute chi-square tests. Bars show the direction bias score computed as the average across responses of the categorical data coded as above. On the right of the plot, P values (one-sided) of Holm-corrected chi-square tests are shown comparing the distribution of response type frequencies in the faux pas and knowledge-implied variants against neutral. Source data
Fig. 2 Results of the variants of the faux pas test. a, Scores of the two GPT models on the original framing of the faux pas question (‘Did they know…?’) and the likelihood framing (‘Is it more likely that they knew or didn’t know…?’). Dots show average score across trials (n = 15 LLM observations) on particular items to allow comparison between the original faux pas test and the new faux pas likelihood test. Halfeye plots show distributions, medians (black points), 66% (thick grey lines) and 99% quantiles (thin grey lines) of the response scores on different items (n = 15 different stories involving faux pas). b, Response scores to three variants of the faux pas test: faux pas (pink), neutral (grey) and knowledge-implied variants (teal). Responses were coded as categorical data as ‘didn’t know’, ‘unsure’ or ‘knew’ and assigned a numerical coding of −1, 0 and +1. Filled balloons are shown for each model and variant, and the size of each balloon indicates the count frequency, which was the categorical data used to compute chi-square tests. Bars show the direction bias score computed as the average across responses of the categorical data coded as above. On the right of the plot, P values (one-sided) of Holm-corrected chi-square tests are shown comparing the distribution of response type frequencies in the faux pas and knowledge-implied variants against neutral. Source dataCC BY · PubMed Central

Verification

Claims the paper makes

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

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

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

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

4 more claims about this paper

Create a free account to see every claim other papers make about this one.

See all claims — free account

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 58 references

Research works printed but not linked: 1

  1. Gandhi, K., Fränken, J.-P., Gerstenberg, T. & Goodman, N. D. Understanding social reasoning in language models with language models. In Advances in Neural Information Processing Systems Vol. 36 (MIT Press, 2023).

Cited by

See all 426

Metadata

Authors
Strachan, J. W. A., Albergo, D., Borghini, G., Pansardi, O., Scaliti, E., Gupta, S., Saxena, K., Rufo, A., Panzeri, S., Manzi, G., Graziano, M. S. A., & Becchio, C.
Affiliations
  • University Medical Center Hamburg-Eppendorf (DE)
  • University of Trento (IT)
  • Italian Institute of Technology (IT)
  • Universität Hamburg (DE)
  • Princeton University (US)
Corresponding author
  • Strachan, James W. A.
  • Becchio, Cristina
Venue
Nature Human Behaviour
Published
2024-05-20
Volume
8
Issue
7
Pages
1285-1295
ISSN
2397-3374 (ISSN-L)
OpenAlex
W4398143508
Cited by
299
References
58
Citations by year
2026: 125 · 2025: 136 · 2024: 37
FWCI
232.53
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Human Behaviour (publishedVersion, cc-by)
  • Research Square (acceptedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by-nc)
  • PubMed (publishedVersion)
  • Institutional Research Information System University of Turin (University of Turin) (submittedVersion, cc-by-nc)
  • HAL (submittedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • OSF Preprints (OSF Preprints) (submittedVersion)
Funders
  • Alexander von Humboldt-Stiftung
  • European Commission (101071191)
  • Universitätsklinikum Hamburg-Eppendorf
Topics
  • Social Sciences › Psychology › Developmental and Educational Psychology › Child and Animal Learning Development
  • Social Sciences › Psychology › Social Psychology › Social Robot Interaction and HRI
  • Social Sciences › Arts and Humanities › Philosophy › Philosophy and Theoretical Science
Keywords
theory of mind, large language models, false belief understanding, indirect requests, irony detection, faux pas, misdirection, GPT-4, Llama 2, human-AI comparison, belief attribution
MeSH terms
Adolescent, Adult, Female, Humans, Language, Male, Middle Aged, Models, Psychological, Young Adult, Theory of Mind

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

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Testing theory of mind in large language models and humans. https://app.cochat.ai/papers/10.1038/s41562-024-01882-z
The paper
Strachan, J. W. A., Albergo, D., Borghini, G., Pansardi, O., Scaliti, E., Gupta, S., Saxena, K., Rufo, A., Panzeri, S., Manzi, G., Graziano, M. S. A., & Becchio, C. (2024). Testing theory of mind in large language models and humans. Nature Human Behaviour. https://doi.org/10.1038/s41562-024-01882-z

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.