as a fun aside here from the abstract it does come down to how you use it: > AI users who maintain similar homework completion time as non-AI users experience

as a fun aside here from the abstract it does come down to how you use it: > AI users who maintain similar homework completion time as non-AI users experience small learning losses. If the pure completion time of homework goes drops significantly with an LLM tool, I suspect these students are spending the extra time turning the material over in different ways to internalize, which is interesting.