I happened to run into Google’s Gemini the other day. I was about to start work on a recommendation letter for an ex-student who was applying for a professor job at a university in Montréal. I was prepping myself to get into that specific kind of writing mode when I saw the little star icon at the top right and hovered it to find out it was Google’s AI.
Immediately, I had a flash forward of me using Gemini to write this recommendation letter based on the document I’d been constructing alongside the ex-student in question. I could have Gemini write it, then edit it, and get the whole thing submitted in… however long that process takes? I don’t actually know. Let’s say 20 minutes? I could be kidding myself.
I didn’t do that, though. I spent an hour or so writing the letter the old fashioned way, with my brain. I felt pretty strongly that I wouldn’t want to receive an AI-written recommendation letter for somebody if I were on a hiring committee, that it would feel like an insult, like their recommender was phoning it in and signalling that they didn’t give too much of a shit about the recommendation. That’s how I tend to feel when I receive student work clearly written by/“with” an LLM, same thing for application letters to our graduate program, and so on.
There are probably exceptions to this – maybe using an LLM helps somebody put together a solid application in a language they aren’t quite strong enough in, despite having excellent ideas and knowledge. And maybe that’s a good thing, definitely could be. I haven’t run into a ton of those cases explicitly, so I haven’t had a chance to talk to anyone in that specific situation.
The thing I was struck by with the Gemini moment was how tempting it was. I feel like I came close and I feel like I dodged a bullet. It reminded me of the compulsion I feel when I play short time control chess games, of just reaching for the sheer satisfaction of this chunk of well-defined time, even when I lose (especially when I lose?). I felt in that moment pretty strongly that if I went down that LLM-writing mode I would be on that road more or less permanently.
It’s that “phoning it in” feeling that stops me mostly, all the things that would be lost in that time saving act. It seems like it would unavoidably be a diminished thing. It’s not just that I think the writing would be trite more or less by definition – hey maybe the writing itself would be fantastic, I really don’t know, though I’ve seen little evidence of it in the wild. It’s more about the loss of the experience of writing, and all the little hooks it has for thought - for thinking about that ex-student and his excellence, to sit with that excellence and to remember conversations with him, things I admire about him, how I would want to put that to make sure the hiring committee would really get it. The world wouldn’t end, obviously, but I’d be losing something precious, assigning it a lower value than I think it truly deserves - and not just for me, but for everyone who receives it as well, like damming a river (okay, a tiny rivulet).
I’m struggling with how to express all this. I’ll try again sometime, because it does seem important. It’s something plenty of people are talking about and trying to address. But ultimately I feel like it’s about making a decision about what matters, and what can and should be phoned in.
I think I’m newly worried, thanks to that tempting moment with Gemini, my evil twin, that we’re going to be fooled by convenience into thinking important things aren’t important; that we won’t be in the right headspace at the right moment to make the decision with the care it deserves.