GENERATIVE AI AND HUMAN THINKING: COGNITIVE OFFLOADING, LEARNING, MEMORY, AND CRITICAL THINKING

Conversational artificial intelligence (AI) systems such as ChatGPT and Claude have rapidly become everyday tools for hundreds of millions of people. Unlike earlier tools, they do not merely retrieve information but analyse, write and draw conclusions on the user's behalf. This narrative review analyses the available empirical evidence on how offloading cognitive tasks to generative AI affects brain engagement, learning, memory and critical thinking. It focuses on an EEG study from the MIT Media Lab (n=54), a survey of 319 knowledge workers by Microsoft Research and Carnegie Mellon University, a randomised field experiment with high-school students and related research. The evidence indicates that uncritical reliance on AI is associated with lower engagement of the user's own cognitive resources during the task, poorer recall of self-written text and a reduced tendency to verify AI output; the higher the trust in AI, the less critical scrutiny is applied. However, the evidence base is still limited: studies are small, some are non-peer-reviewed preprints, and no long-term harm has been demonstrated. A conceptual model of "cognitive debt" accumulation and recommendations for the reasoned use of AI in higher education, including medical education, are proposed.