CITATION — REFERENCE ENTRY
Faster Completion, Less Learning: Generative AI Reduced Study Time on Math Problems and the Knowledge They Build — arXiv
- Key
- rismanchian-et-al-2026-faster-completion-less-learning
- Authors
- Rismanchian, Sina; Uzun, Hasan; Matayoshi, Jeffrey; Cosyn, Eric; Kurd-Misto, Eyad
- Issued
- 2026-5-20
- Type
- article
- Container
- arXiv
Raw CSL JSON
{
"DOI": "10.48550/arXiv.2605.21629",
"URL": "https://arxiv.org/abs/2605.21629",
"type": "article",
"genre": "Preprint",
"title": "Faster Completion, Less Learning: Generative AI Reduced Study Time on Math Problems and the Knowledge They Build",
"author": [
{
"given": "Sina",
"family": "Rismanchian"
},
{
"given": "Hasan",
"family": "Uzun"
},
{
"given": "Jeffrey",
"family": "Matayoshi"
},
{
"given": "Eric",
"family": "Cosyn"
},
{
"given": "Eyad",
"family": "Kurd-Misto"
}
],
"issued": {
"date-parts": [
[
2026,
5,
20
]
]
},
"number": "2605.21629",
"container-title": "arXiv"
}
Claims
-
Using 3.2 million ALEKS learning interactions, the authors found that college students' time on AI-susceptible math problems fell 26.9% over 11 quarters after ChatGPT's release, the divergence vanished under proctoring, and proctored retention showed a 25% cumulative decline in odds of a correct response; they describe this as a population-level indicator of cognitive surrender.
"Learning time on AI-susceptible problems declines 2.8% per quarter among college students after ChatGPT's release, cumulating to 26.9% over eleven quarters [...] Among college students, the post-ChatGPT divergence vanishes entirely under proctoring [...] Logistic fixed-effects models on randomly assigned proctored retention items yield a 25% cumulative decline in odds of correct response [...] the population-level indicator of cognitive surrender"
Available in