Please stop accusing people of not being human!!
Just a quick post tonight, because I’ve been taking part in so many conversations today in different spots online about false accusations of writing being AI-generated. (e.g. this case with an autistic professor, and a case shared by Timnit Gebru on Mastodon)
Reality seems to be proving all of the conjectures I made here back in February about WHOSE writing is likely to be falsely flagged as “AI-generated” to be correct, namely: non-native English writers and the neurodivergent. The former of those, there’s actually published research on now.
The latter, well, I’m still trying to figure out how to ethically, non-exploitatively study whether neurodivergent writing is more likely to flagged as AI-generated, but anecdotally it really does seem to be true. I care very much about doing this research ethically because I do not want to cause harm to a vulnerable community. (The folks who published the article about the bias against L2 writers don’t seem to share that concern; I see no indication anywhere in their paper that they sought specific consent for having their writing fed into an AI-detector from the students whose writing formed the corpus, and every indication that it was simply a corpus of convenience for them). I’m concerned about consent because every AI detector out there is using what’s put into it in some way, and I don’t think it’s safe to assume that just because a text is available, its author necessarily consents to that. (This is why I’ve only ever tested my own writing; if you’ve read my earlier posts, y’all know I’m apparently a robot!)
The other concern I have is that if the results came out the way I suspect they will (showing that ND writers ARE more likely to have their writing flagged as non-human), that could very, very easily be misinterpreted and misused by the folks for whom the narrative that ND folks are somehow less “human” is appealing. There are existing human biases here that I really don’t want to accidentally amplify. We shouldn’t be judging ANYONE’S humanity on the basis of the writing they produce in the first place, and for ND folks, I think the false positives hit harder because of how often we already have been subjected to experiences that make us feel alien, or in which we’re called “robotic”. (Plus, there’s the issue of rejection sensitive dysphoria, which I think also makes these accusations hit harder for us.)
I feel like Cassandra here. People are being hurt by these accusations that at least for me, given the background I have, were so obviously predictable. I’m just so sad and frustrated that for the most part, people STILL aren’t listening. I *am* listening, though, and if you are someone who has been falsely accused of submitting AI generated writing, please feel free to reach out to me. I’ve helped at least one student exonerate themselves, and while I can’t promise I’d be able to do the same for anyone else, I can promise to try.
