UC Irvine study finds ChatGPT use linked to 25% drop in exam scores
A research team at the University of California, Irvine, analyzed 10 years of data from McGraw Hill’s Aleks math platform to assess the impact of ChatGPT on student learning. By comparing student behavior before and after the model’s late-2022 launch, the study found that while AI use increased homework speed and immediate correctness, it correlated with a significant decline in long-term…
Key points
- UC Irvine study of 10 years of Aleks data links ChatGPT use to a 25% drop in exam accuracy.
- Students showed faster homework completion and higher immediate correctness but weaker long-term understanding.
- Middlebury College research found students using AI as a tutor outperformed those using it to draft essays.
The findings suggest that passive reliance on AI may weaken critical thinking skills by allowing students to offload step-by-step reasoning. A separate multi-institutional experiment cited in the study indicated that adults who relied heavily on AI for logic problems became more likely to abandon challenging questions. However, a July study from Middlebury College offers a counterpoint: students who used AI as a tutor rather than a drafting tool achieved higher test scores and essay quality. The research highlights a distinction between using AI to bypass cognitive effort and using it to support learning.
Is ChatGPT Dulling Your Brain? New Studies Raise Alarming Questions.
entrepreneur.com · 1 October 2026
Loading the full article…
This text was published by entrepreneur.com and written by Sherin Shibu. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Introducing Olmo-core 3: Open, scalable training infrastructure for large MoEs · 1 src
- MIT team’s AI beats top Stratego players by 15-1-4 · 2 src
- Qwen4Exp adds MTP support in llama.cpp pull request · 1 src
- Unite.AI outlines stages of AI hallucination and its risks · 1 src
- Deep learning perturbation models outperform baselines under calibrated metrics, study finds · 1 src
Comments
via GitHub Discussions