How to Stay Smart in the Age of AI: The Science of Critical Thinking

Summary

Generative AI is prompting concern that students may outsource the effort through which memory, reasoning and critical-thinking skills develop. Education researcher Lixiang Yan found that the agentic AI Codex could complete assignments, answer questions and participate in discussions for an online course without students being intellectually engaged. A 2025 survey of more than 1,000 US faculty members found that 90% expected generative AI to reduce students’ critical-thinking skills, while about 20% of AI-using higher-education students in a global survey said they already found it harder to solve problems without the technology. Evidence remains unsettled: a 2024 study found that students solved mathematics problems better with ChatGPT-like assistance but performed worse after the tool was removed, while other studies suggest that feedback and thinking prompts can support learning. Researchers emphasize that critical thinking is tied to subject knowledge and must be taught within specific domains through coherent curricula and repeated practice. The article notes that students often prefer quick answers over demanding prompts, and that rigorous long-term evidence is still limited. The Education Endowment Foundation has announced up to £2.5 million for studies of generative AI’s effects on learning, memory and problem-solving, while the OECD is working with Macat on a standardized critical-thinking assessment.