"False"
Skip to content
printicon
Main menu hidden.
Published: 2026-11-13 Updated: 2026-11-19, 11:43

When AI thinks for us: the risk of cognitive surrender

NEWS In this article, Shaw & Nave (2025) introduce their Tri-System Theory, an extension of the Dual Process Theory of thinking. The latter describes human thinking as an interplay between System 1, fast, intuitive thinking, and System 2, slower, analytical thinking. The authors argue that generative AI now constitutes a third system, System 3, consisting of external, artificial cognition.

Text: Lars Larsson

Shaw & Nave examined how AI affects human reasoning in three experiments involving a total of 1,372 participants. The participants were asked to solve tasks that required reflective thinking and could consult an AI assistant whenever they wished. The AI was manipulated so that it sometimes provided correct answers and sometimes convincing but incorrect answers.

A central concept in the article is cognitive surrender. In this context, it refers to situations in which people accept answers from AI with minimal scrutiny and hand over the thinking process to the AI. The results showed that participants often followed the AI’s recommendations. When the AI provided correct answers, performance improved markedly, but when it provided incorrect answers, performance declined considerably. At the same time, participants’ confidence in their answers increased even when the AI led them astray.

The studies also showed that time pressure increased reliance on AI, while financial incentives and feedback made participants more likely to question incorrect AI answers. People with high levels of trust in AI were more susceptible to cognitive surrender, while those with greater analytical ability and a stronger tendency to think critically were more resistant.

Applied to the university context, the findings provide indications that AI functions not only as a tool but also as an active part of both students’ and teachers’ thinking. Shaw & Nave therefore argue that teaching needs to develop students’ ability to critically evaluate AI-generated answers so that AI can be used to support thinking rather than to develop into something that replaces their own judgement.

Reference: Shaw, S. D., & Nave, G. (2025). Thinking—Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning and the Rise of Cognitive Surrender. The Wharton School, University of Pennsylvania.

Read the article