Human Adults and LLMs as Scientists: Who Benefits from Active Exploration?

Forum topic · 小凯 · 2026-06-08

Summary

A long-standing finding in causal learning research is that adults struggle to identify conjunctive causal rules—where an effect requires multiple causes to be present simultaneously—while performing better in disjunctive settings. Most prior demonstrations of this 'conjunctive handicap' relied on passive observation paradigms in which learners could not control evidence generation. This paper, by researchers including Mandana Samiei, Alison Gopnik, and Doina Precup (arXiv:2606.06464), investigates whether the bias persists when adults gain agency through active exploration. Using a modified 'blicket detector' task, participants freely intervened on objects to uncover conjunctive or disjunctive causal rules. The results show that active exploration substantially improves adults' conjunctive causal reasoning, though conjunctive rules still require more tests to infer than disjunctive ones. The authors also compare human performance with a range of large language models on the same task: some state-of-the-art models approach human-level hypothesis inference accuracy, but they tend to employ less efficient exploration strategies and show a similar conjunctive–disjunctive performance gap.

Paper Overview

Field: NLP / Cognitive Science Authors: Mandana Samiei, Eunice Yiu, Anthony GX-Chen, Dongyan Lin, Jocelyn Shen, Blake A. Richards, Alison Gopnik, Doina Precup arXiv: 2606.06464

Abstract (Translation)

A long-standing finding in the causal learning literature is that adults struggle to identify conjunctive causal rules, where an effect requires the simultaneous presence of multiple causes, while performing better in disjunctive settings. However, most demonstrations of this "conjunctive handicap" rely on passive observation paradigms with limited evidence, where learners have no control over evidence generation. This paper asks whether this bias persists when adults are granted agency through active exploration.

Using a modified "blicket detector" task, adult participants freely intervened to identify causal objects under conjunctive or disjunctive rule structures. The study shows that active exploration substantially improves adults' conjunctive causal reasoning, although conjunctive rules still require more tests to infer than disjunctive ones.

The authors further compare human performance with a range of large language models in the same setting. While some state-of-the-art models approach human-level accuracy in hypothesis inference, they tend to exhibit less efficient exploration strategies and a similar conjunctive–disjunctive performance gap.

Key Takeaways

Active exploration mitigates the conjunctive handicap: giving adults agency over evidence generation substantially improves conjunctive causal reasoning compared to passive observation.
Conjunctive rules remain harder: even with active exploration, conjunctive rules require more interventions/tests to infer than disjunctive rules.
LLM comparison: some frontier LLMs match humans on inference accuracy, but their exploration strategies are less efficient, and they show the same conjunctive–disjunctive gap.