Article

Large language models infer hidden preferences from eye movements but fail to exploit response times

2026-07-15

Abstract excerpt

<p>Large language models (LLMs) are increasingly deployed in interactive settings, in which they must inferhuman preferences not only from explicit instructions but also from observed behaviour. To test theirrelevant capabilities, we evaluated three LLMs (Llama 3.1, the cognition-tuned Centaur , and GPT-5 ) asinteraction partners in a multi-round negotiation task where human preferences must be inferred fromchoice...

Topics

Open a Topic to create a Post that cites this publication.

Identifiers and source

Literature Corpus work: ffe0c2a8-60d2-520e-8b25-8dd5a2e3c5ff
DOI: 10.31234/osf.io/bw5kq_v1

Open publication

Related research

Semantic proximity does not establish scientific evidence.

Click a neighbor to travelStep 1 · 12 closest

Large language models infer hidden preferences from eye movements but fail to exploit response timesDOI 10.31234/osf.io/bw5kq_v1Read

Worth the Weight: Modern LLMs Demonstrate Accurate Metacognitive Knowledge of Decision Weights in Multi-Attribute ChoiceSemantic proximity Read

Nonlinear Transformation of Probabilities by Large Language ModelsSemantic proximity Read

Artificial Intelligence and Metacognitive Knowledge: LLMs Can Report Decision Weights as Faithfully as Humans in Multi-Attribute ChoiceSemantic proximity Read

GPT-3.5 altruistic advice is sensitive to reciprocal concerns but not to strategic riskSemantic proximity Read

Assessing Large Language Models’ ability to predict how humans balance self-interest and the interest of othersSemantic proximity Read

Emergent Behavioural Signatures in Large Language Models: A Cross-Task Study of Risk and Forecasting BehaviourSemantic proximity Read

Is ChatGPT Humanly Irrational?Semantic proximity Read

Uncovering the Computational Mechanisms Underlying Many-Alternative ChoiceSemantic proximity Read

Benchmarking Large Language Model Rationality Using Measurement AxiomsSemantic proximity Read

Competitive, Cooperative, and Everything In-Between: The Janus Protocol for Strategy-Infused Language ModelsSemantic proximity Read

Reasoning from social information enables efficient human intelligenceSemantic proximity Read

Centaur: a foundation model of human cognitionSemantic proximity Read