Human Adults and LLMs as Scientists: Active Exploration Reduces the Conjunctive Causal Reasoning Handicap

Forum topic · 小凯 · 2026-06-08

Summary

A well-established finding in causal learning research is that human adults struggle with conjunctive causal rules—cases where an effect requires multiple causes to be present simultaneously—while performing better on disjunctive rules. Most prior demonstrations, however, used passive observation paradigms with limited evidence and no learner control. This paper, from researchers including Mandana Samiei, Alison Gopnik, and Doina Precup (arXiv:2606.06464), investigates whether this "conjunctive handicap" persists when adults can actively explore. Using a modified blicket detector task where participants freely intervened on objects governed by conjunctive or disjunctive rules, the authors show that active exploration substantially improves adults' conjunctive causal reasoning, though conjunctive rules still require more tests to infer. They further compare human performance with a range of large language models in the same setting: while some frontier LLMs approach human-level hypothesis inference accuracy, they tend to use less efficient exploration strategies and exhibit a similar conjunctive-disjunctive performance gap, highlighting differences in how humans and LLMs act as scientists.

Paper Overview

Field: NLP / Causal Reasoning
Authors: Mandana Samiei, Eunice Yiu, Anthony GX-Chen, Dongyan Lin, Jocelyn Shen, Blake A. Richards, Alison Gopnik, Doina Precup
Published: 2026-06-04
arXiv: 2606.06464

Abstract (translated from the Chinese summary)

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 results show 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-vs-disjunctive performance gap.

Key Takeaways

Agency matters: Allowing learners to actively generate evidence (intervene) significantly reduces the classic conjunctive reasoning deficit seen in passive observation paradigms.
A residual gap remains: Conjunctive rules are still harder to infer than disjunctive ones, even with free exploration.
LLM comparison: Frontier LLMs can match humans on final hypothesis inference, but their exploration strategies are less efficient, and they show a comparable conjunctive handicap—suggesting shared structural biases but different exploration behavior.

Original Abstract (excerpt)

> 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. We show that active exploration substantially improves adults' conjunctive causal reasoning, although conjunctive rules ...

*Auto-collected on 2026-06-08*