Does relying on AI reduce users' critical thinking?
It can: knowledge workers who trusted GenAI more reported less critical thinking, and developers using an AI assistant wrote less secure code while feeling more confident.
Early signal, tested on real LLM products
Treat it as a hypothesis.
A large self-report survey and one small controlled study, both with real GenAI tools.
2 studies · 2 with LLMs · 2023–2025 · How we rate
Design AI help so it keeps users checking, especially where they over-trust it.
- Prompt verification at the points where users are most likely to over-trust.From the study
- Build for users' own confidence: people with more self-confidence thought more critically.From the study
- Flag security-sensitive or high-stakes output for review.From the study
- Support verification and integration, which is where critical thinking moved.From the study
What they found
Lee et al. (2025)survey
Survey of 319 knowledge workers who shared 936 real examples of using GenAI at work. Higher confidence in GenAI was associated with less critical thinking, and higher self-confidence with more. Critical thinking shifted toward verifying information, integrating responses and stewarding the task. Self-reported and correlational.
“higher confidence in GenAI is associated with less critical thinking, while higher self-confidence is associated with more critical thinking”
Lee et al., 2025 · Abstract
Perry et al. (2023)experiment
47 participants did five security-related programming tasks in Python, JavaScript and C, with or without an AI code assistant. Those with the assistant wrote insecure solutions more often on four of the five tasks, and were more likely to believe their code was secure.
“Participants with access to an AI assistant were also more likely to believe they wrote secure code, suggesting that such tools may lead users to be overconfident about security flaws in their code.”
Perry et al., 2023 · Abstract
Trade-offs and when it doesn't apply
- Lee et al. is self-reported and correlational.
- Perry et al. had 47 participants.
- Some offloading is healthy and saves time.
What it looks like
How to check it in your product
Track how often users edit or question AI output over time; a steady fall may signal over-reliance.
Open questions
- Does long-term AI use erode skills or just shift them?
- Which nudges keep critical thinking without slowing users?
- How do novices learn when AI does the hard parts?
FAQ
Does trusting AI reduce critical thinking?
In Lee et al.'s survey of 319 knowledge workers, higher confidence in GenAI was associated with less critical thinking.
Do AI coding assistants make code less secure?
In Perry et al. (47 participants), those with an assistant wrote insecure solutions more often on four of five tasks and were more likely to think their code was secure.
Sources
- The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge WorkersHao-Ping (Hank) Lee, Advait Sarkar, Lev Tankelevitch, et al. (2025). CHI 2025.
- Do Users Write More Insecure Code with AI Assistants?Neil Perry, Megha Srivastava, Deepak Kumar, Dan Boneh (2023). CCS 2023.