Reasoning Across Minds and Machines
Abstract
Reasoning is one of the hallmarks of both natural and artificial intelligence. Understanding how reasoning operates in the human mind is crucial in cognitive science. Despite a long history of research on human reasoning in cognitive science—ranging from heuristics (Tversky & Kahneman, 1974) to mental models (Johnson-Laird, 1983), and from Bayesian modeling (Oaksford & Chater, 2007; Griffiths, Chater, & Tenenbaum, 2024) to neuroscience (Goel & Dolan, 2003)—little is known about how humans reason so flexibly in real life and how reasoning contributes to high-level cognitive functions including planning, social interaction, complex problem-solving, and open-ended mental exploration. Previous studies face challenges that hinder a deeper understanding of reasoning, including the difficulty of designing well-balanced experimental paradigms that maintain both control and ecological validity, efficient data collection and analysis beyond pure behavioral measures (e.g., Think-Aloud text data (Simon & Ericsson, 1984)), and understanding complex interactions between reasoning and other high-level cognitive functions, such as memory, theory of mind, and language.