English

Eliciting the Priors of Large Language Models using Iterated In-Context Learning

Computation and Language 2024-06-05 v1

Abstract

As Large Language Models (LLMs) are increasingly deployed in real-world settings, understanding the knowledge they implicitly use when making decisions is critical. One way to capture this knowledge is in the form of Bayesian prior distributions. We develop a prompt-based workflow for eliciting prior distributions from LLMs. Our approach is based on iterated learning, a Markov chain Monte Carlo method in which successive inferences are chained in a way that supports sampling from the prior distribution. We validated our method in settings where iterated learning has previously been used to estimate the priors of human participants -- causal learning, proportion estimation, and predicting everyday quantities. We found that priors elicited from GPT-4 qualitatively align with human priors in these settings. We then used the same method to elicit priors from GPT-4 for a variety of speculative events, such as the timing of the development of superhuman AI.

Keywords

Cite

@article{arxiv.2406.01860,
  title  = {Eliciting the Priors of Large Language Models using Iterated In-Context Learning},
  author = {Jian-Qiao Zhu and Thomas L. Griffiths},
  journal= {arXiv preprint arXiv:2406.01860},
  year   = {2024}
}
R2 v1 2026-06-28T16:52:10.922Z