I’m looking at the first reviews we received tagged with “use of generative AI in the review.” If and how much AI was used is actually not clear, but if so, they are stirring my worst fears.
There are (at least) two things wrong with the reviews we received.
Hallucination
One reviewer is completely wrong. Our paper is about companies selling program code to each other and how to price that and the reviewer argues that this is covered by the employment relationship at a company, ignoring the inter-company aspect. This suggests the reviewer didn’t read the paper and leaves open where their comments are coming from, presumably an AI then.
An AI not getting an innovation seems apt, after all it is only an innovation if it is not the AI’s training data, or only minimally in the training data. So reviewers really shouldn’t use a generative AI for anything but touching up their language.
However, even just improving the writing is a bad idea.
Feedback
I’m trying to understand the feedback and it presents itself as a mostly impentrable wall of weasel words. The reviewers “provide”, “suggest”, “incentivize”, etc. They “commend on our valuable contribution” and “wish us good luck in future endeavors”. Can we please get no-bullshit speak back into reviews?
Seriously, in my research course, I teach clear writing with a vengeance. I tell my students to avoid all hand-wavy language and to be precise. AIs, however, seem to be primed to use vague and imprecise wording, for reasons unknown to me. Parts of the reviews read like those well-written student letters I get inundated with that are all the same and have obviously been written by an AI to hide the author’s lack of command of the English language.
Summary
Please, don’t use AIs in writing peer reviews.
Leave a Reply