AI Use Raises Concerns About Declining Thinking Ability

AI is now handling everything from wedding speeches, tax filing, and war trauma counseling, absorbing not just memory and calculation but also analysis, idea generation, and decision-making.

Anthropic's analysis of over 80,000 interviews revealed both the benefits of using AI as a learning tool and a dependency where people stop thinking for themselves. Workers in law, finance, government, and healthcare reported the greatest practical gains from AI, but also had a high rate of directly experiencing errors. Students, teachers, and academics reported both learning benefits and concerns about cognitive decline together, while tech workers reported strong learning benefits with almost no concerns about decline. However, since long-term tracking data is lacking, the conclusions remain tentative for now.

Timing also matters. When time is scarce, using AI from the start improved performance, but when there was sufficient time, early use reduced memory retention and narrowed thinking, lowering performance. Conversely, thinking for oneself first and then using AI led to deeper consideration of opposing viewpoints and broader answers.

• Uncritical acceptance can lead to cognitive surrender.
• Expert competence is built through practice, and automating too early can weaken that process.
• Some researchers argue that tools merely change incentives, and evidence that humans' fundamental abilities themselves are collapsing remains weak.
• The key is learning to use AI as a metacognitive partner rather than a substitute.

What matters in work and learning is not how much you delegate to AI, but when you delegate and when you think for yourself.