[carbine]

I skimmed this but am I reading correctly, participants were given 20 minutes to write an essay and asked to do their best and then given (or not given) access to a tool to help? There's zero incentive here not to optimize for shortcuts and task completion.

This is very different from, say, writing an essay I'm gonna publish on my blog under my own name. I would be MUCH more interested in an experiment that isolates people working on highly cognitively demanding work that MATTERS to them, and seeing what impact LLMs do (or don't) have on cognitive function. Otherwise, this seems like a study designed to confirm a narrative.

What am I missing