"very likely" Means "uncertain"? How LLMs Diverge from Humans in Linguistic Uncertainty Quantification

Jinhao Duan, Zicheng Liu, Zijie Liu, Kaidi Xu, Tianlong Chen
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:26965-26989, 2026.

Abstract

Humans express uncertainty verbally via markers (e.g., "possible", "likely"), yet most LLM uncertainty quantification (UQ) relies on costing likelihood- or consistency-based signals. From a cognitive perspective, accurate verbal uncertainty reflects metacognitive monitoring, representing knowledge boundaries ("knowing that you don’t know") to support regulation and information seeking. In this paper, we investigate: How LLMs diverge from humans in verbal uncertainty quantification? Can verbal markers reliably quantify LLM uncertainty? We curate a corpus of human uncertainty markers from psychology and decision-science literature and benchmark LLMs against it. We observe that LLMs encode verbal uncertainty with numerical levels that differ substantially from those of humans. We then introduce VOCAL, a novel optimization-based algorithm that learns an optimal uncertainty profile over uncertainty markers directly from LLM outputs. By fitting a marker–uncertainty mapping to best explain empirical correctness, VOCALdiscovers how much probability mass each verbal marker should convey, rather than estimating uncertainty via repeated sampling. VOCAL enables a direct, marker-level comparison of confidence semantics between humans and LLMs, disentangling mismatch and revealing systematic confidence disparities in verbal expressions.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-duan26b, title = {"very likely" Means "uncertain"? {H}ow {LLM}s Diverge from Humans in Linguistic Uncertainty Quantification}, author = {Duan, Jinhao and Liu, Zicheng and Liu, Zijie and Xu, Kaidi and Chen, Tianlong}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {26965--26989}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/duan26b/duan26b.pdf}, url = {https://proceedings.mlr.press/v306/duan26b.html}, abstract = {Humans express uncertainty verbally via markers (e.g., "possible", "likely"), yet most LLM uncertainty quantification (UQ) relies on costing likelihood- or consistency-based signals. From a cognitive perspective, accurate verbal uncertainty reflects metacognitive monitoring, representing knowledge boundaries ("knowing that you don’t know") to support regulation and information seeking. In this paper, we investigate: How LLMs diverge from humans in verbal uncertainty quantification? Can verbal markers reliably quantify LLM uncertainty? We curate a corpus of human uncertainty markers from psychology and decision-science literature and benchmark LLMs against it. We observe that LLMs encode verbal uncertainty with numerical levels that differ substantially from those of humans. We then introduce VOCAL, a novel optimization-based algorithm that learns an optimal uncertainty profile over uncertainty markers directly from LLM outputs. By fitting a marker–uncertainty mapping to best explain empirical correctness, VOCALdiscovers how much probability mass each verbal marker should convey, rather than estimating uncertainty via repeated sampling. VOCAL enables a direct, marker-level comparison of confidence semantics between humans and LLMs, disentangling mismatch and revealing systematic confidence disparities in verbal expressions.} }
Endnote
%0 Conference Paper %T "very likely" Means "uncertain"? How LLMs Diverge from Humans in Linguistic Uncertainty Quantification %A Jinhao Duan %A Zicheng Liu %A Zijie Liu %A Kaidi Xu %A Tianlong Chen %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-duan26b %I PMLR %P 26965--26989 %U https://proceedings.mlr.press/v306/duan26b.html %V 306 %X Humans express uncertainty verbally via markers (e.g., "possible", "likely"), yet most LLM uncertainty quantification (UQ) relies on costing likelihood- or consistency-based signals. From a cognitive perspective, accurate verbal uncertainty reflects metacognitive monitoring, representing knowledge boundaries ("knowing that you don’t know") to support regulation and information seeking. In this paper, we investigate: How LLMs diverge from humans in verbal uncertainty quantification? Can verbal markers reliably quantify LLM uncertainty? We curate a corpus of human uncertainty markers from psychology and decision-science literature and benchmark LLMs against it. We observe that LLMs encode verbal uncertainty with numerical levels that differ substantially from those of humans. We then introduce VOCAL, a novel optimization-based algorithm that learns an optimal uncertainty profile over uncertainty markers directly from LLM outputs. By fitting a marker–uncertainty mapping to best explain empirical correctness, VOCALdiscovers how much probability mass each verbal marker should convey, rather than estimating uncertainty via repeated sampling. VOCAL enables a direct, marker-level comparison of confidence semantics between humans and LLMs, disentangling mismatch and revealing systematic confidence disparities in verbal expressions.
APA
Duan, J., Liu, Z., Liu, Z., Xu, K. & Chen, T.. (2026). "very likely" Means "uncertain"? How LLMs Diverge from Humans in Linguistic Uncertainty Quantification. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:26965-26989 Available from https://proceedings.mlr.press/v306/duan26b.html.

Related Material