The Impact of Generative AI on Critical Thinking: Cognitive Scaffolding, Substitution, and the Risk of Deskilling

Abstract The rapid diffusion of generative artificial intelligence (AI) has raised fundamental questions about its effects on critical thinking and cognitive development. Evidence from recent empirical studies suggests a dual pattern: structured forms of AI assistance can sustain users’ cognitive activity and enhance engagement and performance, whereas more substitutive forms of use are associated with reduced autonomous reasoning and poorer subsequent independent performance. This article argues that these findings can be understood by considering how cognitive work is distributed between users and AI systems and how this distribution may evolve over time. In particular, I examine a six-step sequence in which successful complementary assistance can increase trust and delegation; greater delegation can remain complementary or become increasingly substitutive; and increasingly substitutive interactions can reduce the autonomous exercise of critical thinking and potentially hinder the development or maintenance of critical-thinking capacities over time. The available evidence provides substantial support for the first five steps considered individually and substantial, though less direct, support for the sixth. The article concludes by discussing educational, design, and social implications, arguing that preserving cognitive agency requires fostering complementary rather than progressively substitutive forms of interaction.