Language Agents Mirror Human Causal Reasoning Biases. How Can We Help Them Think Like Scientists?

2025-05-14

AI Summary

This paper identifies a systematic “disjunctive bias” in large language models (LMs) for causal reasoning: LMs reliably infer common disjunctive causal relationships but significantly underperform on well-supported yet uncommon conjunctive ones. This bias is robust across model architectures, scales, and prompting strategies, intensifies with task complexity, and closely mirrors adult human cognitive biases—suggesting LMs internalize heuristic patterns from training data. To characterize it empirically, the authors adapt the Blicket test paradigm to construct a behavioral evaluation framework, providing the first formal definition and experimental validation of this bias. They further propose a test-time intervention—multi-hypothesis sampling coupled with logical elimination—that mitigates the bias without parameter modification. Results show substantial gains in conjunctive causal identification accuracy, advancing LMs toward scientifically rigorous causal modeling. (149 words)

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Language model (LM) agents are increasingly used as autonomous decision-makers who need to actively gather information to guide their decisions. A crucial cognitive skill for such agents is the efficient exploration and understanding of the causal structure of the world -- key to robust, scientifically grounded reasoning. Yet, it remains unclear whether LMs possess this capability or exhibit systematic biases leading to erroneous conclusions. In this work, we examine LMs' ability to explore and infer causal relationships, using the well-established"Blicket Test"paradigm from developmental psychology. We find that LMs reliably infer the common, intuitive disjunctive causal relationships but systematically struggle with the unusual, yet equally (or sometimes even more) evidenced conjunctive ones. This"disjunctive bias"persists across model families, sizes, and prompting strategies, and performance further declines as task complexity increases. Interestingly, an analogous bias appears in human adults, suggesting that LMs may have inherited deep-seated reasoning heuristics from their training data. To this end, we quantify similarities between LMs and humans, finding that LMs exhibit adult-like inference profiles (but not children-like). Finally, we propose a test-time sampling method which explicitly samples and eliminates hypotheses about causal relationships from the LM. This scalable approach significantly reduces the disjunctive bias and moves LMs closer to the goal of scientific, causally rigorous reasoning.

Problem

• LMs struggle with conjunctive causal relationships
• LMs exhibit disjunctive bias similar to humans
• Proposing method to reduce bias in causal reasoning

Innovation

• LM agents explore causal relationships using Blicket Test
• Test-time sampling method reduces disjunctive bias
• Scalable approach enhances scientific causal reasoning