Cognitive offloading: the more we trust AI, the less we think

What the study examined

A joint research team from Carnegie Mellon University and Microsoft Research surveyed 319 knowledge workers to understand how generative AI use relates to perceived competence and to the amount of cognitive effort people invest in their work.

Key finding

The more people trust generative AI tools, the less mental effort they are willing to invest in solving a task themselves.

Reduced cognitive effort

The research confirmed a strong correlation between high confidence in AI output and a conscious or unconscious reduction of one own cognitive effort.

The false confidence effect

Users tend to accept AI answers without deeper reflection. The speed and fluency of generated text creates a false sense that the task has been solved correctly.

Cognitive laziness

Once a tool is easily available, the mind naturally optimises energy and delegates even tasks that previously required deep analytical thinking.

What to do about it

Do not outsource responsibility for thinking. If you delegate the entire framing and structuring phase to AI, your brain gradually loses the ability to solve complex problems on its own.

• Spend five minutes brainstorming on paper before you write a prompt.
• Formulate your own hypothesis before you ask the model.
• Use AI as a sparring partner to review your ideas, not as their source.

Source

Lee, H.-P., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. (2025). The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers. CHI 2025. doi.org/10.1145/3706598.3713778

This is one of the six areas measured by the AI Cognitive Health Radar.