Two posts on the same preprint (the preprint is behind a paywall

Two posts on the same preprint (the preprint is behind a paywall as far as I can see, which is a first for me) about a large-scale study of students using AI in China. Long story short: 1️⃣ According to Gabriel Demombynes AI bad, no AI good. The figure here seems to support that conclusion. 2️⃣ A much more nuanced, and in my view insightful, take from Ethan Mollick (yes, Ethan is often more nuanced than people assume): "using AI hurts learning if it undermines mental effort. When homework time drops due to AI use, so do test scores. Students who spent the same amount of time studying did not have learning loss." That's the problem with reporting only part of the study. It reinforces the widespread opinion that "AI is hurting learning" whereas in fact it depends critically on how AI is used. As Mollick writes, "the instinct of students is often to use AI to help with homework, even if they are not trying to cheat. And because off-the-shelf chatbots are a helpful assistant, rather than a tutor, they give you the answer and undermine learning. AI tutors, on the other hand, challenge learners and lead to increases in test scores." It is crucial to not make overly broad statements about "using AI." Depending on how it is used, it can be useful or not, harmful or not, create long-term learning or cognitive surrender. By dismissing it entirely as a danger to learning, or as a distraction from productivity, we are not focusing enough on what can be done to improve outcomes. To paraphrase the amazing Darren Walker, nuance could use some allies.