Using AI chatbots for even just for 10 minutes may have a shockingly negative impact on people’s ability to think and problem-solve, according to a new study from researchers at Carnegie Mellon, MIT, Oxford, and UCLA.
-
@GossiTheDog not gonna lie: this work is kinda problematic, especially when I see how they frame their results.
The first study simply represents a modular change problem. If you had help you will always struggle initially if that help is cut off suddenly, and the ability only slowly recovers.
The second, "better" version (acc to authors) is also funny. Once you read "Participants who used AI for hints showed no significant impairments relative to control", it's easy to guess what happened. Lazy people did not take the task seriously and used AI for answers. Of course they'll crash.
This rather confirms my personal notion that AI is mostly enabling laziness rather than killing thinking ability (as I stated here some time ago: https://k4tana.github.io/blog/2025/06/04/AI-Nuanced-Take.html).Edit: Additionally those effect sizes are not really much to write home about, especially in the second experiment. Statistical significance alone does not mean practical relevance.
@odr_k4tana @GossiTheDog I really would have liked to see another group who was only given the test questions without the prior "practice" questions. That could have shown whether AI use directly decreased performance or if it merely made the practice less effective
-
Using AI chatbots for even just for 10 minutes may have a shockingly negative impact on people’s ability to think and problem-solve, according to a new study from researchers at Carnegie Mellon, MIT, Oxford, and UCLA.
https://ai-project-website.github.io/AI-assistance-reduces-persistence/
@GossiTheDog
"In the AI condition, participants solved 12 fraction problems with an AI assistant (GPT-5) available in a sidebar. The AI was then removed without warning, and all participants solved 3 additional test problems independently. Participants in the AI condition had a significantly lower solve rate"I'm quite sure the same would happen with a calculator instead of an AI chatbot. Interesting study but with this methodology I'd take it with a grain of salt.
-
P pelle@veganism.social shared this topic
