CITATION — REFERENCE ENTRY
AI Meets the Classroom: When Do Large Language Models Harm Learning? — arXiv
- Key
- lehmann-2024-ai-meets-classroom
- Authors
- Lehmann, Matthias; Cornelius, Philipp B.; Sting, Fabian J.
- Issued
- 2024-8-29
- Type
- article
- Publisher
- arXiv
Raw CSL JSON
{
"DOI": "10.48550/arXiv.2409.09047",
"URL": "https://arxiv.org/abs/2409.09047",
"type": "article",
"genre": "Preprint",
"title": "AI Meets the Classroom: When Do Large Language Models Harm Learning?",
"author": [
{
"given": "Matthias",
"family": "Lehmann"
},
{
"given": "Philipp B.",
"family": "Cornelius"
},
{
"given": "Fabian J.",
"family": "Sting"
}
],
"issued": {
"date-parts": [
[
2024,
8,
29
]
]
},
"number": "arXiv:2409.09047",
"publisher": "arXiv"
}
Claims
-
Two preregistered laboratory experiments on novice Python learners found no effect of LLM access on overall learning outcomes.
"In two pre-registered and incentivized laboratory experiments, we find no effect of LLMs on overall learning outcomes."
-
LLMs widened the gap in learning between students with low and high prior knowledge.
"We also observe that LLMs widen the gap between students with low and high prior knowledge."
-
In exploratory analyses, students who used LLMs to generate solutions covered more topics but understood each less, while students who asked LLMs for explanations did not cover more topics but understood them better.
"Students who substitute some of their learning activities with LLMs (e.g., by generating solutions to exercises) increase the volume of topics they can learn about but decrease their understanding of each topic. Students who complement their learning activities with LLMs (e.g., by asking for explanations) do not increase topic volume but do increase their understanding."
Available in