The debate over understanding in AI’s large language models
Open accessThe DOI matches its Crossref record; 78 of the 78 references it cites were found in a registry or an index.
Abstract
We survey a current, heated debate in the artificial intelligence (AI) research community on whether large pretrained language models can be said to understand language—and the physical and social situations language encodes—in any humanlike sense. We describe arguments that have been made for and against such understanding and key questions for the broader sciences of intelligence that have arisen in light of these arguments. We contend that an extended science of intelligence can be developed that will provide insight into distinct modes of understanding, their strengths and limitations, and the challenge of integrating diverse forms of cognition.
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.
Integrity signals
Registry
Cites
Statistics
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.
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?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“While a growing number of studies have shown that LLMs exhibit an inherent brittleness (13) and unreliability (14, 15) in knowledge tasks, and that they struggle with real-world understanding (11, 16) and human-like reasoning tasks (9, 10, 14), others have shown that LLMs instead excel in mimicking human language and communicative abilities (17)”
From The benefits and dangers of anthropomorphic conversational agents (2025) · cites this paper as [11]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“In their work, an argument is built up towards the inability of a machine to attain human level recognition, with the important distinction between form and meaning being drawn.”
From Deanthropomorphising NLP: Can a language model be conscious? (2024) · cites this paper as [84]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“This concern is often raised under the guise of the ELIZA effect: a phenomenon in which participants attribute human characteristics to even very simple systems [27,28].”
From People cannot distinguish GPT-4 from a human in a Turing test (2025) · cites this paper as [28]
The cited source says what the claim attributes to it.
In the citing paper
“Instead of resorting to human metaphors, it may be more productive to consider LLMs as a new type of 'understanding' distinct from human cognition (Mitchell & Krakauer, 2023).”
From Using ChatGPT and Other Large Language Model (LLM) Applications for Academic Paper Assignments (2023) · cites this paper as Mitchell & Krakauer, 2023
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
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805 · Cited by 73,173 · cited 4×
“An LLM called BERT [16] obtained near-human performance on this benchmark [38].”
BackgroundInfluentialHighly accurate protein structure prediction with AlphaFold
Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., Tunyasuvunakool, K., Bates, R., Žídek, A., Potapenko, A., et al. (2021) · Nature · doi:10.1038/s41586-021-03819-2 · Cited by 43,764 · cited 1×
“…“understanding”, one that enables extraordinary, superhuman predictive ability, such as in the case of the AlphaZero and AlphaFold systems from DeepMind [40, 72], which respectively seem to bring an “alien” form of intuition to the domains of chess playing and protein-structure prediction [39, 68].”
BackgroundLanguage Models are Few-Shot Learners
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.14165 · Cited by 23,372 · cited 5×
“Those on the other side of this debate argue that large pretrained models such as GPT-3 or LaMDA—however fluent their linguistic output—cannot possess understanding because they have no experience or mental models of the world; their training in predicting words in vast collections of text has taught them the form of l…”
BackgroundMethodInfluentialKaugeranna, R., Kaugeranna, E., & Aion, (. S. 4. 6. (2023) · DROPS (Schloss Dagstuhl – Leibniz Center for Informatics) · doi:10.4230/lipics.giscience.2023.43 · Cited by 13,690
On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜
Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021) · Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency · doi:10.1145/3442188.3445922 · Cited by 6,988
Barsalou, L. W. (2008) · Annual Review of Psychology · doi:10.1146/annurev.psych.59.103006.093639 · Cited by 5,936 · cited 1×
“Researchers disagree on the extent to which concepts are domain-specific and innate versus more general-purpose and learned [14, 30, 31, 53, 75, 85], the degree to which concepts are grounded via embodied metaphors [28, 46, 61], are represented in the brain via dynamic, situation-based simulations [5], and the conditi…”
BackgroundChain-Of-Thought Prompting Elicits Reasoning in Large Language Models
Wei, J., Wang, X., Schuurmans, D., Bosma, M., Ichter, B., Xia, F., Chi, E., Le, Q. V., & Zhou, D. (2022) · Advances in Neural Information Processing Systems 35 · doi:10.52202/068431-1800 · Cited by 4,505
Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
Wei, J., Wang, X., Schuurmans, D., Bosma, M., Ichter, B., Xia, F., Chi, E., Le, Q., & Zhou, D. (2022) · arXiv · doi:10.48550/arxiv.2201.11903 · Cited by 4,263 · cited 1×
“LLMs such as OpenAI’s well-known GPT-3 (5) and more recent ChatGPT (6) and Google’s PaLM (7) can produce astonishingly humanlike text, conversation, and, in some cases, what seems like human reasoning abilities (8), even though the models were not explicitly trained to reason.”
BackgroundGLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding
Wang, A., Singh, A., Michael, J., Hill, F., Levy, O., & Bowman, S. (2018) · Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP · doi:10.18653/v1/w18-5446 · Cited by 4,222 · cited 2×
“For example, two standard benchmarks for assessing LLMs are the General Language Understanding Evaluation (GLUE) [46], and its successor (SuperGLUE) [47], which include large-scale datasets with tasks such as “textual entailment” (given two sentences, can the meaning of the second be inferred from the first?), “words…”
BackgroundTraining language models to follow instructions with human feedback
Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C. L., Mishkin, P., Zhang, C., Agarwal, S., Slama, K., Ray, A., et al. (2022) · arXiv · doi:10.48550/arxiv.2203.02155 · Cited by 4,208 · cited 1×
“Such models can be used to generate natural language, be fine-tuned for specific language tasks [58], or be further trained to better match “user intent” [65].”
Background
Research works printed but not linked: 3
- 10.1037/0033-295X.99.4.587 · doi:10.1037/0033-295x.99.4.587 · Paper page
- R. W. Gibbs Metaphor Wars (Cambridge University Press 2017).
- 10.1207/s15516709cog0402_4 · doi:10.1207/s15516709cog0402_4 · Paper page
Cited by
Artificial intelligence and illusions of understanding in scientific research
Messeri, L., & Crockett, M. J. (2024) · Nature · doi:10.1038/s41586-024-07146-0 · Cited by 707
“Although the question of what AI models can ‘understand’ remains controversial 14 , the AI visions (alongside the language of ‘deep learning’, ‘deep neural networks’ and marketing hype 95 ) portray AI as having a depth of understanding beyond what can be grasped by limited human minds.”
Can Generative AI improve social science?
Bail, C. A. (2024) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2314021121 · Cited by 369
Can Generative AI Improve Social Science?
Bail, C. A. (2023) · doi:10.31235/osf.io/rwtzs · Cited by 369
Dissociating language and thought in large language models
Mahowald, K., Ivanova, A. A., Blank, I. A., Kanwisher, N., Tenenbaum, J. B., & Fedorenko, E. (2024) · Trends in Cognitive Sciences · doi:10.1016/j.tics.2024.01.011 · Cited by 345
Susarla, A., Gopal, R., Thatcher, J. B., & Sarker, S. (2023) · Information Systems Research · doi:10.1287/isre.2023.ed.v34.n2 · Cited by 285
Metadata
- Authors
- Mitchell, M., & Krakauer, D. C.
- Affiliations
- Santa Fe Institute (US)
- Corresponding author
- Krakauer, David C.
- Venue
- Proceedings of the National Academy of Sciences
- Publisher
- National Academy of Sciences
- Published
- 2023-03-21
- Volume
- 120
- Issue
- 13
- Pages
- e2215907120-e2215907120
- ISSN
- 0027-8424 (ISSN-L)
- 1091-6490
- PMC
- PMC10068812
- OpenAlex
- W4328049044
- Cited by
- 340
- References
- 56
- Citations by year
- 2026: 107 · 2025: 105 · 2024: 84 · 2023: 42
- FWCI
- 31.82
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- Locations
- Proceedings of the National Academy of Sciences (publishedVersion, cc-by-nc-nd)
- PubMed Central (publishedVersion, cc-by-nc-nd)
- PubMed (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- National Science Foundation (2020103)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Social Sciences › Arts and Humanities › Philosophy › Philosophy and Theoretical Science
- Social Sciences › Arts and Humanities › History and Philosophy of Science › Cybernetics and Technology in Society
- Keywords
- large language models, language comprehension, artificial intelligence, cognition, pre-trained language models
- MeSH terms
- Artificial Intelligence, Cognition, Dissent and Disputes
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: The debate over understanding in AI’s large language models. https://app.cochat.ai/papers/10.1073/pnas.2215907120
- The paper
- Mitchell, M., & Krakauer, D. C. (2023). The debate over understanding in AI’s large language models. Proceedings of the National Academy of Sciences. https://doi.org/10.1073/pnas.2215907120
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.