AI assistance study finds weaker performance after help is removed

What's this about?

Three randomized trials with 1,222 participants tested fractions and reading comprehension. Brief AI access improved assisted performance, but accuracy and persistence fell after removal. The updated paper does not establish permanent cognitive damage or the same effect in every professional workflow.

What exactly is randomized trial?

Participants are assigned by chance to conditions, helping separate an intervention’s effect from pre-existing differences.

What's the community saying?

1. Should the trial also compare calculators, textbooks or human assistance?
2. Does withdrawal measure a temporary adjustment, rather than lasting loss of ability?
3. Users report both dependence and higher motivation when AI removes technical roadblocks.