Do Large Language Models Show Biases in Causal Learning? Insights from Contingency Judgment

TL;DR

This work investigates whether large language models exhibit the illusion of causality by adapting the classic contingency judgment task to LLM prompts in health-focused scenarios. Using a dataset of 1,000 null-contingency medical cases, and three contemporary LLMs (GPT-4o-Mini, Claude-3.5-Sonnet, Gemini-1.5-Pro), the authors assess whether models infer causality where evidence is lacking. Nonparametric tests reveal robust model-specific biases: all models produce unwarranted causal judgments, with substantial variability across models and higher likelihoods of zero-causality responses in some cases; these results support the view that LLMs reproduce causal language rather than manifest true causal understanding. The study highlights important implications for deploying LLMs in domains requiring precise causal inference and suggests directions for improving evaluation methods, prompting strategies, and data collection to mitigate such biases.

Abstract

Causal learning is the cognitive process of developing the capability of making causal inferences based on available information, often guided by normative principles. This process is prone to errors and biases, such as the illusion of causality, in which people perceive a causal relationship between two variables despite lacking supporting evidence. This cognitive bias has been proposed to underlie many societal problems, including social prejudice, stereotype formation, misinformation, and superstitious thinking. In this work, we examine whether large language models are prone to developing causal illusions when faced with a classic cognitive science paradigm: the contingency judgment task. To investigate this, we constructed a dataset of 1,000 null contingency scenarios (in which the available information is not sufficient to establish a causal relationship between variables) within medical contexts and prompted LLMs to evaluate the effectiveness of potential causes. Our findings show that all evaluated models systematically inferred unwarranted causal relationships, revealing a strong susceptibility to the illusion of causality. While there is ongoing debate about whether LLMs genuinely understand causality or merely reproduce causal language without true comprehension, our findings support the latter hypothesis and raise concerns about the use of language models in domains where accurate causal reasoning is essential for informed decision-making.

Figures (4)

• Figure 1: Distribution of outputs across models in null-contingency scenarios.
• Figure 2: Models’ responses across the four variable categories.
• Figure 3: Results generated under deterministic conditions (temperature = 0), with one sample per prompt.
• Figure 4: Results under their default temperature setting, with one sample per prompt.