Theory Is All You Need: AI, Human Cognition, and Causal Reasoning.

This article critically examines the differences between artificial intelligence (AI) and human cognition, arguing that AI’s data-driven, prediction-based approach fundamentally differs from human theory-based causal reasoning. While AI systems, such as large language models (LLMs), excel at processing vast amounts of past data to generate fluent outputs, they lack the forward-looking capacity to generate genuine novelty, new knowledge, or causal understanding necessary for decision making under uncertainty. The authors illustrate this distinction through examples like language learning and the historical case of heavier-than-air flight, emphasizing that human cognition involves data–belief asymmetries where beliefs can precede and motivate the generation of new data via experimentation. They conclude that AI’s reliance on past data and statistical associations limits its ability to replace human decision making in uncertain and novel contexts, highlighting opportunities for future research on human–AI collaboration, task-specific capabilities, and the theoretical foundations of cognition.