{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,7]],"date-time":"2026-09-07T09:12:38Z","timestamp":1788772358272,"version":"build-2803163510"},"reference-count":50,"publisher":"Springer Science and Business Media LLC","license":[{"start":{"date-parts":[[2026,9,7]],"date-time":"2026-09-07T00:00:00Z","timestamp":1788739200000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by-nc-nd\/4.0"},{"start":{"date-parts":[[2026,9,7]],"date-time":"2026-09-07T00:00:00Z","timestamp":1788739200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by-nc-nd\/4.0"}],"funder":[{"name":"Google DeepMind"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["Nat Mach Intell"],"DOI":"10.1038\/s42256-026-01293-x","type":"journal-article","created":{"date-parts":[[2026,9,7]],"date-time":"2026-09-07T09:02:26Z","timestamp":1788771746000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Causal evidence that language models use confidence to drive behaviour"],"prefix":"10.1038","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-5995-5264","authenticated-orcid":false,"given":"Dharshan","family":"Kumaran","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5029-1430","authenticated-orcid":false,"given":"Nathaniel","family":"Daw","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Simon","family":"Osindero","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2820-4692","authenticated-orcid":false,"given":"Petar","family":"Veli\u010dkovi\u0107","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Viorica","family":"Patraucean","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"297","published-online":{"date-parts":[[2026,9,7]]},"reference":[{"key":"1293_CR1","doi-asserted-by":"publisher","first-page":"366","DOI":"10.1038\/nn.4240","volume":"19","author":"A Pouget","year":"2016","unstructured":"Pouget, A., Drugowitsch, J. & Kepecs, A. Confidence and certainty: distinct probabilistic quantities for different goals. Nat. Neurosci. 19, 366\u2013374 (2016).","journal-title":"Nat. Neurosci."},{"key":"1293_CR2","doi-asserted-by":"publisher","first-page":"1322","DOI":"10.1098\/rstb.2012.0037","volume":"367","author":"A Kepecs","year":"2012","unstructured":"Kepecs, A. & Mainen, Z. F. A computational framework for the study of confidence in humans and animals. Philos. Trans. R. Soc. B 367, 1322\u20131337 (2012).","journal-title":"Philos. Trans. R. Soc. B"},{"key":"1293_CR3","doi-asserted-by":"crossref","unstructured":"Steyvers, M. & Peters, M. A. K. Metacognition and uncertainty communication in humans and large language models. Curr. Dir. Psychol. Sci. 35, 131\u2013139 (2026).","DOI":"10.1177\/09637214251391158"},{"key":"1293_CR4","doi-asserted-by":"publisher","first-page":"419","DOI":"10.1016\/j.tics.2022.02.004","volume":"26","author":"C Stone","year":"2022","unstructured":"Stone, C., Mattingley, J. B. & Rangelov, D. On second thoughts: changes of mind in decision-making. Trends Cogn. Sci. 26, 419\u2013431 (2022).","journal-title":"Trends Cogn. Sci."},{"key":"1293_CR5","doi-asserted-by":"publisher","first-page":"614","DOI":"10.1038\/s42256-026-01217-9","volume":"8","author":"D Kumaran","year":"2026","unstructured":"Kumaran, D. et al. Competing biases underlie overconfidence and underconfidence in LLMs. Nat. Mach. Intell. 8, 614\u2013627 (2026).","journal-title":"Nat. Mach. Intell."},{"key":"1293_CR6","unstructured":"Xiong, M. et al. Can LLMs express their uncertainty? An empirical evaluation of confidence elicitation in LLMs. In International Conference on Learning Representations (eds Kim, B. et al.) 23650\u201323678 (Curran Associates, 2023)."},{"key":"1293_CR7","doi-asserted-by":"crossref","unstructured":"Tian, K. et al. Just ask for calibration: strategies for eliciting calibrated confidence scores from language models fine-tuned with human feedback. In Proc. 2023 Conference on Empirical Methods in Natural Language Processing (eds Bouamor, H. et al.) 5433\u20135442 (Association for Computational Linguistics, 2023).","DOI":"10.18653\/v1\/2023.emnlp-main.330"},{"key":"1293_CR8","doi-asserted-by":"publisher","first-page":"221","DOI":"10.1038\/s42256-024-00976-7","volume":"7","author":"M Steyvers","year":"2025","unstructured":"Steyvers, M. et al. What large language models know and what people think they know. Nat. Mach. Intell. 7, 221\u2013231 (2025).","journal-title":"Nat. Mach. Intell."},{"key":"1293_CR9","doi-asserted-by":"publisher","first-page":"91","DOI":"10.1037\/rev0000045","volume":"124","author":"SM Fleming","year":"2017","unstructured":"Fleming, S. M. & Daw, N. D. Self-evaluation of decision-making: a general Bayesian framework for metacognitive computation. Psychol. Rev. 124, 91\u2013114 (2017).","journal-title":"Psychol. Rev."},{"key":"1293_CR10","doi-asserted-by":"publisher","first-page":"227","DOI":"10.1038\/nature07200","volume":"455","author":"A Kepecs","year":"2008","unstructured":"Kepecs, A., Uchida, N., Zariwala, H. A. & Mainen, Z. F. Neural correlates, computation and behavioural impact of decision confidence. Nature 455, 227\u2013231 (2008).","journal-title":"Nature"},{"key":"1293_CR11","doi-asserted-by":"publisher","first-page":"551","DOI":"10.1016\/j.cub.2007.01.061","volume":"17","author":"AL Foote","year":"2007","unstructured":"Foote, A. L. & Crystal, J. D. Metacognition in the rat. Curr. Biol. 17, 551\u2013555 (2007).","journal-title":"Curr. Biol."},{"key":"1293_CR12","doi-asserted-by":"publisher","first-page":"759","DOI":"10.1126\/science.1169405","volume":"324","author":"R Kiani","year":"2009","unstructured":"Kiani, R. & Shadlen, M. N. Representation of confidence associated with a decision by neurons in the parietal cortex. Science 324, 759\u2013764 (2009).","journal-title":"Science"},{"key":"1293_CR13","doi-asserted-by":"publisher","first-page":"529","DOI":"10.1162\/tacl_a_00754","volume":"13","author":"B Wen","year":"2025","unstructured":"Wen, B. et al. Know your limits: a survey of abstention in large language models. Trans. Assoc. Comput. Linguist. 13, 529\u2013556 (2025).","journal-title":"Trans. Assoc. Comput. Linguist."},{"key":"1293_CR14","doi-asserted-by":"crossref","unstructured":"Kirichenko, P., Ibrahim, M., Chaudhuri, K. & Bell, S. J. AbstentionBench: reasoning LLMs fail on unanswerable questions. In Advances in Neural Information Processing Systems (eds Belgrave, D. et al.) (Neural Information Processing Systems Foundation, 2026).","DOI":"10.52202\/085713-5729"},{"key":"1293_CR15","unstructured":"Madhusudhan, N., Madhusudhan, S. T., Yadav, V. & Hashemi, M. Do LLMs know when to not answer? Investigating abstention abilities of large language models. In Proc. 31st International Conference on Computational Linguistics (eds Rambow, O. et al.) 10137\u201310153 (Association for Computational Linguistics, 2025)."},{"key":"1293_CR16","doi-asserted-by":"publisher","first-page":"136037","DOI":"10.52202\/079017-4322","volume":"37","author":"A Arditi","year":"2024","unstructured":"Arditi, A. et al. Refusal in language models is mediated by a single direction. Adv. Neural Inf. Process. Syst. 37, 136037\u2013136083 (2024).","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"1293_CR17","unstructured":"Chuang, Y.-N. et al. Learning to route LLMs with confidence tokens. Preprint at https:\/\/arxiv.org\/abs\/2410.13284 (2024)."},{"key":"1293_CR18","unstructured":"Tjandra, B. A., Razzak, M., Kossen, J., Handa, K. & Gal, Y. Fine-tuning large language models to appropriately abstain with semantic entropy. Preprint at https:\/\/arxiv.org\/abs\/2410.17234 (2024)."},{"key":"1293_CR19","doi-asserted-by":"crossref","unstructured":"Zhang, H. et al. R-tuning: instructing large language models to say \u2018I don\u2019t know\u2019. In Proc. 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers) (eds Duh, K. et al.) 7113\u20137139 (Association for Computational Linguistics, 2024).","DOI":"10.18653\/v1\/2024.naacl-long.394"},{"key":"1293_CR20","unstructured":"Plaut, B., Khanh, N. X. & Trinh, T. Probabilities of chat LLMs are miscalibrated but still predict correctness on multiple-choice Q&A. Preprint at https:\/\/arxiv.org\/abs\/2402.13213 (2024)."},{"key":"1293_CR21","unstructured":"Yadkori, Y. A. et al. Mitigating LLM hallucinations via conformal abstention. Preprint at https:\/\/arxiv.org\/abs\/2405.01563 (2024)."},{"key":"1293_CR22","unstructured":"Tomani, C., Chaudhuri, K., Evtimov, I., Cremers, D. & Ibrahim, M. Uncertainty-based abstention in LLMs improves safety and reduces hallucinations. Preprint at https:\/\/arxiv.org\/abs\/2404.10960 (2024)."},{"key":"1293_CR23","doi-asserted-by":"publisher","first-page":"535","DOI":"10.1146\/annurev.neuro.29.051605.113038","volume":"30","author":"JI Gold","year":"2007","unstructured":"Gold, J. I. & Shadlen, M. N. The neural basis of decision making. Annu. Rev. Neurosci. 30, 535\u2013574 (2007).","journal-title":"Annu. Rev. Neurosci."},{"key":"1293_CR24","doi-asserted-by":"publisher","first-page":"1310","DOI":"10.1098\/rstb.2011.0416","volume":"367","author":"N Yeung","year":"2012","unstructured":"Yeung, N. & Summerfield, C. Metacognition in human decision-making: confidence and error monitoring. Philos. Trans. R. Soc. B 367, 1310\u20131321 (2012).","journal-title":"Philos. Trans. R. Soc. B"},{"key":"1293_CR25","unstructured":"Guo, C., Pleiss, G., Sun, Y. & Weinberger, K. Q. On calibration of modern neural networks. In International Conference on Machine Learning 1321\u20131330 (PMLR, 2017)."},{"key":"1293_CR26","first-page":"9459","volume":"33","author":"P Lewis","year":"2020","unstructured":"Lewis, P. et al. Retrieval-augmented generation for knowledge-intensive NLP tasks. Adv. Neural Inf. Process. Syst. 33, 9459\u20139474 (2020).","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"1293_CR27","unstructured":"Bommasani, R. et al. On the opportunities and risks of foundation models. Preprint at https:\/\/arxiv.org\/abs\/2108.07258 (2021)."},{"key":"1293_CR28","doi-asserted-by":"crossref","unstructured":"Reimers, N. & Gurevych, I. Sentence-BERT: sentence embeddings using Siamese BERT-networks. In Proc. 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (eds Inui, K. et al.) 3982\u20133992 (Association for Computational Linguistics, 2019).","DOI":"10.18653\/v1\/D19-1410"},{"key":"1293_CR29","unstructured":"Turner, A. M. et al. Steering language models with activation engineering. Preprint at https:\/\/arxiv.org\/abs\/2308.10248 (2023)."},{"key":"1293_CR30","doi-asserted-by":"crossref","unstructured":"Stolfo, A., Balachandran, V., Yousefi, S., Horvitz, E. & Nushi, B. Improving instruction-following in language models through activation steering. In International Conference on Learning Representations 55790\u201355823 (Association for Computational Linguistics, 2025).","DOI":"10.32388\/BTHT4K"},{"key":"1293_CR31","doi-asserted-by":"crossref","unstructured":"VanderWeele, T. Explanation in Causal Inference: Methods for Mediation and Interaction (Oxford Univ. Press, 2015).","DOI":"10.1093\/ije\/dyw277"},{"key":"1293_CR32","doi-asserted-by":"crossref","unstructured":"Wen, B., Howe, B. & Wang, L. L. Characterizing LLM abstention behavior in science QA with context perturbations. In Findings of the Association for Computational Linguistics: EMNLP 2024 (eds Al-Onaizan, Y. et al.) 3437\u20133450 (Association for Computational Linguistics, 2024).","DOI":"10.18653\/v1\/2024.findings-emnlp.197"},{"key":"1293_CR33","unstructured":"Sclar, M., Choi, Y., Tsvetkov, Y. & Suhr, A. Quantifying language models\u2019 sensitivity to spurious features in prompt design or: how I learned to start worrying about prompt formatting. In International Conference on Learning Representations 25055\u201325083 (Curran Associates, 2024)."},{"key":"1293_CR34","unstructured":"Geng, J. et al. A survey of language model confidence estimation and calibration. Preprint at https:\/\/arxiv.org\/abs\/2311.08298 (2023)."},{"key":"1293_CR35","unstructured":"Yoon, D. et al. Reasoning models better express their confidence. In Advances in Neural Information Processing Systems (eds Belgrave, D. et al.) 103869\u2013103896 (Neural Information Processing Systems Foundation, 2026)."},{"key":"1293_CR36","unstructured":"Steyvers, M., Belem, C. & Smyth, P. Improving metacognition and uncertainty communication in language models. Preprint at https:\/\/arxiv.org\/abs\/2510.05126 (2025)."},{"key":"1293_CR37","unstructured":"Kumaran, D. et al. How do LLMs compute verbal confidence. In Proc. 43rd International Conference on Machine Learning (eds Emerson, T. et al.) (PMLR, 2026)."},{"key":"1293_CR38","doi-asserted-by":"crossref","unstructured":"Niculescu-Mizil, A. & Caruana, R. Predicting good probabilities with supervised learning. In Proc. 22nd International Conference on Machine Learning 625\u2013632 (ACM, 2005).","DOI":"10.1145\/1102351.1102430"},{"key":"1293_CR39","unstructured":"Kumaran, D. Reported confidence in LLMs tracks commitment more than correctness. Preprint at https:\/\/arxiv.org\/abs\/2606.29490 (2026)."},{"key":"1293_CR40","doi-asserted-by":"crossref","unstructured":"Rimsky, N. et al. Steering llama 2 via contrastive activation addition. In Proc. 62nd Annual Meeting of the Association for Computational Linguistics 15504\u201315522 (Association for Computational Linguistics, 2024).","DOI":"10.18653\/v1\/2024.acl-long.828"},{"key":"1293_CR41","unstructured":"Kumaran, D., Patraucean, V., Osindero, S., Veli\u010dkovi\u0107, P. & Daw, N. How LLMs detect and correct their own errors: the role of internal confidence signals. Preprint at https:\/\/arxiv.org\/abs\/2604.22271 (2026)."},{"key":"1293_CR42","unstructured":"Gandhi, K., Chakravarthy, A., Singh, A., Lile, N. & Goodman, N. D. Cognitive behaviors that enable self-improving reasoners, or, four habits of highly effective STaRs. Preprint at https:\/\/arxiv.org\/abs\/2503.01307 (2025)."},{"key":"1293_CR43","unstructured":"Venhoff, C., Arcuschin, I., Torr, P., Conmy, A. & Nanda, N. Understanding reasoning in thinking language models via steering vectors. Preprint at https:\/\/arxiv.org\/abs\/2506.18167 (2025)."},{"key":"1293_CR44","unstructured":"Tao, L. et al. Revisiting uncertainty estimation and calibration of large language models. Preprint at https:\/\/arxiv.org\/abs\/2505.23854 (2025)."},{"key":"1293_CR45","unstructured":"Kadavath, S. et al. Language models (mostly) know what they know. Preprint at https:\/\/arxiv.org\/abs\/2207.05221 (2022)."},{"key":"1293_CR46","unstructured":"Lindsey, J. Emergent introspective awareness in large language models (Anthropic, 2025); https:\/\/transformer-circuits.pub\/2025\/introspection\/"},{"key":"1293_CR47","unstructured":"Wei, J. et al. Measuring short-form factuality in large language models. Preprint at https:\/\/arxiv.org\/abs\/2411.04368 (2024)."},{"key":"1293_CR48","unstructured":"Hua, T. T., Qin, A., Marks, S. & Nanda, N. Steering evaluation-aware language models to act like they are deployed. Preprint at https:\/\/arxiv.org\/abs\/2510.20487 (2025)."},{"key":"1293_CR49","doi-asserted-by":"publisher","unstructured":"Kumaran, D. Code and data for Causal Abstention paper. Open Science Framework https:\/\/doi.org\/10.17605\/OSF.IO\/RESW4 (2026).","DOI":"10.17605\/OSF.IO\/RESW4"},{"key":"1293_CR50","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-023-39737-2","volume":"14","author":"TW Webb","year":"2023","unstructured":"Webb, T. W., Miyoshi, K., So, T. Y., Rajananda, S. & Lau, H. Natural statistics support a rational account of confidence biases. Nat. Commun. 14, 3992 (2023).","journal-title":"Nat. Commun."}],"container-title":["Nature Machine Intelligence"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.nature.com\/articles\/s42256-026-01293-x.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s42256-026-01293-x","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s42256-026-01293-x.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,7]],"date-time":"2026-09-07T09:02:48Z","timestamp":1788771768000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.nature.com\/articles\/s42256-026-01293-x"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,7]]},"references-count":50,"alternative-id":["1293"],"URL":"https:\/\/doi.org\/10.1038\/s42256-026-01293-x","relation":{},"ISSN":["2522-5839"],"issn-type":[{"value":"2522-5839","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,7]]},"assertion":[{"value":"12 December 2025","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"20 July 2026","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"7 September 2026","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"The authors declare no competing interests","order":1,"name":"Ethics","label":"Competing interests","group":{"name":"EthicsHeading","label":"Ethics"}}]}}