Journal article · 2026 · Open access

Competing Biases underlie Overconfidence and Underconfidence in LLMs

Dharshan Kumaran, Stephen M. Fleming, Larisa Markeeva, Joe Heyward, Andrea Banino, Mrinal Mathur, Razvan Pascanu, Simon Osindero, Benedetto De Martino, Petar Veličković, Viorica Pătrăucean

Nature Machine Intelligence, 8(4), 614–627 · Published 22 April 2026 · DOI 10.1038/s42256-026-01217-9

Open PDFView at the publisherCite this paper

In brief

“Competing Biases underlie Overconfidence and Underconfidence in LLMs” is a 2026 journal article by Dharshan Kumaran and 10 co-authors, published in Nature Machine Intelligence. OpenAlex counts 46 citations as of 17 November 2026, which puts it in the top 3% of works published in 2026 by citations. It is open access under a CC BY-NC-ND licence: the full text is free to read at Nature Machine Intelligence.

Citations
46
Most cited in
2026 (41)
Field-weighted impact
36.27
References
38
Authors
11

Abstract

Large language models (LLMs) are increasingly deployed in high-stakes applications where reliable confidence estimation is crucial for trustworthy artificial intelligence (AI). However, their confidence dynamics remain poorly understood, with users reporting paradoxical behaviours: LLMs exhibit reduced flexibility in updating initial responses while simultaneously showing excessive sensitivity to contradictory feedback. Understanding these confidence patterns is essential for developing more reliable AI systems and improving human–AI interaction. Here we show that LLM confidence is governed by two competing mechanisms that explain this paradox. First, we identify a choice-supportive bias: when LLMs view their initial answers, they exhibit inflated confidence and maintain their original responses at rates exceeding optimal decision-making, even when presented with contrary evidence. Second, we demonstrate systematic overweighting of contradictory information: LLMs update their confidence more strongly in response to opposing advice than supporting advice, deviating markedly from optimal Bayesian reasoning. These mechanisms operate across diverse models and generalize from simple factual queries to reasoning tasks. Our computational modelling reveals that these two principles—self-consistency preservation and hypersensitivity to contradiction—capture LLM behaviour across domains. These findings provide an understanding of when and why LLMs exhibit adherence to initial responses versus disproportionate updating, with implications for enhancing the robustness and transparency of LLM decision-making. Kumaran et al. show that large language model (LLM) confidence is shaped by two competing biases: a choice-supportive bias that inflates confidence in initial answers, and a systematic overweighting of contradictory advice, deviating from optimal Bayesian reasoning.

Cite this paper

Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.

APA 7

Kumaran, D., Fleming, S. M., Markeeva, L., Heyward, J., Banino, A., Mathur, M., Pascanu, R., Osindero, S., De Martino, B., Veličković, P., & Pătrăucean, V. (2026). Competing Biases underlie Overconfidence and Underconfidence in LLMs. Nature Machine Intelligence, 8(4), 614–627. https://doi.org/10.1038/s42256-026-01217-9

MLA 9

Kumaran, Dharshan, et al. “Competing Biases underlie Overconfidence and Underconfidence in LLMs.” Nature Machine Intelligence, vol. 8, no. 4, 2026, pp. 614–627. https://doi.org/10.1038/s42256-026-01217-9.

Chicago

Kumaran, Dharshan, Stephen M. Fleming, Larisa Markeeva, Joe Heyward, Andrea Banino, Mrinal Mathur, Razvan Pascanu, et al. 2026. “Competing Biases underlie Overconfidence and Underconfidence in LLMs.” Nature Machine Intelligence 8 (4): 614–627. https://doi.org/10.1038/s42256-026-01217-9.

Harvard

Kumaran, D. et al. (2026) ‘Competing Biases underlie Overconfidence and Underconfidence in LLMs’, Nature Machine Intelligence, 8(4), pp. 614–627. doi:10.1038/s42256-026-01217-9.

IEEE

D. Kumaran et al., “Competing Biases underlie Overconfidence and Underconfidence in LLMs,” Nature Machine Intelligence, vol. 8, no. 4, pp. 614–627, 2026, doi: 10.1038/s42256-026-01217-9.

BibTeX

@article{kumaran2026competing,
  title     = {{Competing Biases underlie Overconfidence and Underconfidence in LLMs}},
  author    = {Kumaran, Dharshan and Fleming, Stephen M. and Markeeva, Larisa and Heyward, Joe and Banino, Andrea and Mathur, Mrinal and Pascanu, Razvan and Osindero, Simon and De Martino, Benedetto and Veličković, Petar and Pătrăucean, Viorica},
  journal   = {Nature Machine Intelligence},
  year      = {2026},
  volume    = {8},
  number    = {4},
  pages     = {614--627},
  doi       = {10.1038/s42256-026-01217-9},
  url       = {https://doi.org/10.1038/s42256-026-01217-9}
}

Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)

Topic and keywords

Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling

  • large language models
  • confidence estimation
  • overconfidence
  • underconfidence
  • Bayesian updating
  • cognitive biases
  • human-AI interaction

Work with this paper

Most-cited papers in Computer Science

All highly cited Computer Science papers

Questions about this paper

How do I cite this paper in APA style?

Kumaran, D., Fleming, S. M., Markeeva, L., Heyward, J., Banino, A., Mathur, M., Pascanu, R., Osindero, S., De Martino, B., Veličković, P., & Pătrăucean, V. (2026). Competing Biases underlie Overconfidence and Underconfidence in LLMs. Nature Machine Intelligence, 8(4), 614–627. https://doi.org/10.1038/s42256-026-01217-9

How many times has this paper been cited?

46 times, according to OpenAlex on 17 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.

Is this paper free to read?

Yes. It is open access at Nature Machine Intelligence: sign in and open the PDF from this page.

Who wrote this paper?

Dharshan Kumaran, Stephen M. Fleming, Larisa Markeeva, Joe Heyward, Andrea Banino, Mrinal Mathur, Razvan Pascanu, Simon Osindero, Benedetto De Martino, Petar Veličković, Viorica Pătrăucean.

What is the DOI of this paper?

10.1038/s42256-026-01217-9, permanently at https://doi.org/10.1038/s42256-026-01217-9. Sign in and open it from this page.

Bibliographic data from OpenAlex (CC0), last checked 17 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.