Large language models often see false patterns more than humans do
Illusory Pattern Perception Drives Spurious Inference in Large Language Models
Artificial Intelligence
Summary
Humans sometimes mistakenly find meaningful connections in random information, called illusory pattern perception. The authors studied whether large language models (LLMs) have the same tendency and found that these models can actually see false patterns more often than people. For example, LLMs might wrongly link popular positive traits with big groups or invent causes from unclear events. The authors used a special method to peek inside the models and found features that explain why these mistakes happen. This suggests that LLM reasoning can sometimes be less reliable than expected.
What this means in practice
- •For ai system designers: Improve robustness by detecting and mitigating false pattern recognition in language model outputs to reduce reasoning errors in applications.
- •For bias mitigation teams: Identify and address over-generalizations linking positive attributes to groups or organizations in LLMs to enhance fairness and accuracy.
Authors
Peihua Mai, Zhuoyan Shao, Xinbao Qiao, Meng Zhang, Xinyue Zhou, Yan Pang
Abstract
Illusory pattern perception is a well-documented human cognitive tendency to infer meaningful relationships in data that is actually random. Such a tendency, often described as "connecting the dots" where none exist, can result in systematic reasoning errors. This paper investigates whether Large Language Models (LLMs) exhibit such perceptual tendencies, which can lead to systematic errors in downstream applications. To our knowledge, this work presents the first systematic study of illusory pattern perception in LLMs, adapting classic psychological paradigms to three tasks with direct empirical comparison to human behaviors. We find that LLMs frequently exhibit stronger illusory pattern perception than humans. In particular, models tend to over-associate frequent positive attributes with majority groups or large organizations, and show increased tendencies to construct causal narratives from ambiguous events. To uncover the mechanism behind these behaviors, we develop a feature interpretability framework based on Sparse Autoencoders (SAEs) to analyze internal representations. Our results reveal that holistic frequency perception and analytic cognitive orientation are linked to the emergence of illusory perceptions. These findings highlight a previously underexplored cognitive-like illusion that may affect the reliability of LLM reasoning. Code available at https://github.com/NusIoraPrivacy/illusory.