You open a chat window before you open your own thoughts. The answer lands in seconds and reads well, so you paste it into your work. Michael Gerlich of SBS Swiss Business School set out to measure what that routine costs.
Gerlich surveyed 666 valid respondents in the United Kingdom and interviewed 50 of them. The 23-question survey measured AI tool use, cognitive offloading (handing memory and problem-solving to a device) and critical thinking, with items drawn from the Halpern Critical Thinking Assessment and Terenzini’s self-report measures.
The three correlations point one way. AI use and offloading rose together (r = 0.72). AI use and critical thinking moved apart (r = -0.68). Offloading and critical thinking showed the strongest inverse link (r = -0.75). A mediation analysis put numbers on the bridge. The total effect of AI use on critical thinking was b = -0.42. The path through offloading carried -0.25 of it, about 60 percent, and the direct effect held at -0.17. Offloading carried most of the link between AI use and weaker critical thinking.
Age and schooling shaped the pattern. Participants aged 17 to 25 used AI the most and scored lowest on critical thinking, while those aged 46 and older used it least and scored highest. Interviewees with advanced degrees described cross-checking AI answers against other references. Several with less schooling told Gerlich they follow what the tool suggests. One interviewee put the cost in plain terms, “the less I feel the need to problem-solve on my own.”
The design limits the claim. Gerlich names two limits himself, reliance on self-reported measures and potential sample bias, and the recruitment ran through social media on a convenience basis. A correlation cannot assign cause. Either heavy AI use erodes critical thinking, or weaker critical thinkers lean on AI harder. Gerlich lists experiments as the next step. The paper also notes that offloading frees working capacity and does not by definition reduce engagement.
Two habits follow from the findings, and both are untested suggestions. Write your own two-sentence answer before you open the chat, then compare the two. Check one claim in the reply against a second reference. The cross-check matches what the interviewees with advanced degrees described, and the paper recommends active learning and metacognitive training to counter offloading.
Tomorrow, write your own answer before the chat window gets one.
Reference
Gerlich, M. Correction: Gerlich, M. AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies 2025, 15, 6. Societies 2025, 15, 252. https://doi.org/10.3390/soc15090252
Top comments (0)