"AI brain rot" and cognitive offloading Metacognition: knowing what you know
The more people trusted the AI, the less critical thinking they reported doing
Researchers asked 319 working adults who use generative AI every week to describe real tasks they had done with it. When people were more confident the AI could do a task, they were less likely to report thinking critically about it; when they were more confident in their own ability, they were more likely to. Overall, people felt AI made critical thinking take less effort, and the thinking shifted toward checking and fitting together what the AI produced.
| Paper | The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers |
| Authors | Hao-Ping (Hank) Lee, Advait Sarkar, Lev Tankelevitch, Ian Drosos, Sean Rintel, Richard Banks, Nicholas Wilson |
| Published | 25 April 2025 · CHI ’25: Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems (ACM) |
| Study type | Online survey with regression models and qualitative analysis · Peer-reviewed (full conference paper) |
| What we read | Full text read. We read the full paper (Microsoft Research PDF of the CHI ’25 version). |
| Summarized |
What they did
An online survey on Prolific. Each participant described up to three real work tasks where they used a generative AI tool (936 examples in total), said whether they thought critically during it and why, and rated the effort of six thinking activities from Bloom's taxonomy with and without AI. The team fitted one logistic and six linear regression models with a random intercept per participant, corrected for multiple comparisons, and coded the free-text answers.
Who took part
319 knowledge workers (after excluding 14 of 333 low-quality responses) who said they used generative AI at work at least once a week. 86 were aged 18–24 and 143 were 25–34. Adults at work, not students or children.
What they found
- Participants said they used critical thinking in 555 of the 936 examples (59.29%).
- Overall, they felt the thinking activities took less effort when using a generative AI tool than without one.
- Higher confidence in the AI's ability to do a task was associated with less critical thinking.
- Higher confidence in their own ability to do the task without AI was associated with more critical thinking.
- In their descriptions, critical thinking moved from producing work toward verifying information, integrating the AI's responses and overseeing the task.
Effect size: Regression coefficients are reported in the paper's tables per activity; the headline findings are the direction of the confidence effects, significant after Benjamini–Hochberg correction.
Limitations
- Everything is self-reported, and the authors note participants sometimes confused less effort using the tool with less effort thinking.
- Self-confidence may not match actual skill.
- English-speaking sample, skewed younger and more tech-savvy, all weekly AI users.
- Cross-sectional: it shows associations, not that trusting AI reduces critical thinking.
- Most authors work at Microsoft Research, whose company sells generative AI tools.
What this means for learners and parents
- For learners, the takeaway is calibration: the moments you feel surest the AI is right are the moments you are least likely to check.
- Building a child's own confidence in a skill before handing it to AI may keep them checking, not just accepting.
- A quick habit helps: estimate the answer first, then check the AI's answer against it.
Put it into practice (free tools)
Citation
Hao-Ping (Hank) Lee, Advait Sarkar, Lev Tankelevitch, Ian Drosos, Sean Rintel, Richard Banks, Nicholas Wilson (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 ’25: Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems (ACM). https://doi.org/10.1145/3706598.3713778
Related summaries
- Unrestricted ChatGPT help raised practice scores but lowered exam scores · Hamsa Bastani et al., 2025 · Preregistered randomized controlled trial
- "Your Brain on ChatGPT": what the MIT essay-writing study did and didn't show · Nataliya Kosmyna et al., 2025 · Lab experiment with EEG
- ChatGPT made research feel easier but students' reasoning was weaker than with Google · Matthias Stadler et al., 2024 · Randomized experiment
- With AI, people scored better on reasoning puzzles but overestimated how well they did · Daniela Fernandes et al., 2026 · Two online experiments