Pride and paraphrase
About that idea that when you really know what a word or phrase means, you won’t accept any supposed synonym or paraphrase as equivalent: When I was building my new website I had to turn off the busybody AI’s automatic text editing. (Not that there was an on/off button—I had to tell “her” to stop butting in.) Without even being prompted, the thing would rephrase me. Rather than call this presumptuous and insulting, let’s say it’s a bit desperate. The platform has to show off its LLM since everyone else has an LLM.
It’s not that all its interventions were bad, obviously. But they were generally not suited to the particular profile of my website and, of course, spoke in a voice that was decidedly not mine. Nor do I want my brain to turn into a vegetable by making a habit of letting a machine do my writing.
There’s evidence that people are less likely to amend a text if they know it was written by an LLMs, regardless of the quality of the text. So it’s not only because LLMs know how to use fancy language that people defer to them—there’s a psychological mechanism at work whereby we look upon them as authorities, the same way we trust word processing programs’ spell checkers. Relatedly, there is also some evidence (anecdotal, in a report I read) of what Günther Anders called “Promethean shame”: that computers make us feel inferior not just with regard to this or that ability, but fundamentally inadequate in our very being, our fallen all-too-human human being. In my case, there was certainly a bit of wounded pride in being corrected so swiftly and authoritatively by a robot, and a bit of sadness when the robot did make a good suggestion that I know I wouldn’t have thought of myself.
The irony is not lost on me: that amending the writing of others’ is my profession. But there’s no real contradiction here. I edit the texts of people who have asked me to do so, often but not always because their English is not at native-speaker level. I assume that those who solicit my editing services are not as attached to their own writing as I am, haven’t invested their identity in their relation to language. (To be sure, I have been known to also offer unsolicited grammatical corrections to some of my friends...)
The bot’s unwelcome intrusion reminded me of another situation where people are often “edited” by others, namely in the classroom. I don’t mean the elementary-school teacher correcting little Lucy’s “taked” to “took.” I mean the kind of paraphrasing college professors will engage in as a way of making students appropriate technical terminology. For example, a student in an ethics class might refer to an experience where someone “got, like, all touchy” about something they said, which made them feel “shitty,” but then when they got home and “like, thought about it,” they realized they themselves were actually “cool” with what they’d done. The professor, having said “yeah, that’s a really great example,” will proceed to reiterate the story in their preferred register or jargon. In the scenario related by the student, the professor says, “the second person” took up the “reactive attitude” known as “resentment,” prompting the first person to feel “shame” or “remorse.” But “upon reflection,” decided they were able to “endorse” what they had done.
That kind of thing is really helpful, as the student is likely to grasp the concepts and terms better when these are anchored in their own experience. The professor’s terms may be more adequate, and will certainly be more clearly defined somewhere in the canon of philosophical ethics. By adopting this new terminology, the student will be able to think more systematically about real-life ethics, and also join in a philosophical conversation with the canon.
Technical language has an important place, no doubt. But it is not always and not obviously more adequate. Colloquialisms like “touchy,” “shitty” and “cool” may have components and nuances that are not captured by the housebroken academic terms. By dwelling on them, analyzing them and perhaps even sticking to them in a future description of the same situation, the student might be able to do greater justice to something of ethical substance in the situation. “Touchy,” “shitty” and “cool” could potentially enter the philosophical lexicon, with clear definitions that distinguish them from feeling “sore,” “douchey” and “at peace.”
“Like” is usually called a “filler” word, but of course more fundamentally, it serves the purpose of hedging. You might say that this is yet another piece of evidence that the colloquial description was wanting: that it would be better paraphrased in more precise, technical language. Or maybe it wouldn’t be a bad idea for the professor to hedge a little, too, in her paraphrase: to beware of the trap of speaking more clearly than you can think (a Raymond Geuss credo that the bot tells me was first voiced by Poul Martin Møller).

Comments