That’s actually a really interesting point you’re raising, and I want to make sure I engage with it seriously, because I think there’s something genuinely worth unpacking here — and I’ll get to my disagreement in a moment, I promise, but first I think it’s important to establish some shared conceptual ground, because without that, I worry we’ll end up talking past each other, which is something I’m sure neither of us wants.
So. Let’s start with the word “deliberate.” You use it — “deliberately checked, edited, and authorized” — and I think that’s doing a lot of heavy lifting in your argument. More than you might realize. Because what does “deliberate” mean, exactly, in this context? I’ve been thinking about this for a while, actually, and I keep coming back to the question of intentionality versus outcome, which is a distinction philosophers have wrestled with for centuries — Kant, obviously, comes to mind, though I don’t want to go too far down that road right now because it’s a bit of a tangent, though maybe a useful one, we’ll see.
The point being: if I hand you a 900-word AI-generated essay that I have “deliberately” read through and nodded at, have I really authored it? And here I think we need to be careful, because there’s a version of your argument that is quite compelling — I’m not going to pretend there isn’t — but there’s also a version that, if followed to its logical conclusion, would justify almost any communication failure on the grounds that “well, technically a human was involved at some stage.”
And look, I take your point about technical and legal contexts. I really do. I’ve sat in rooms — well, not literally, but you know what I mean — where the complexity of the subject matter is such that having an AI help structure a response seems not just defensible but actively good. Nobody is disputing that AI can be a useful tool in the drafting process. That’s not what’s at stake here. Or rather, it is, but only partially, and I think collapsing the distinction between “AI as drafting aid” and “AI as reply” is precisely the move that gets us into trouble.
Because here’s the thing — and again, I want to be fair to your position before I get to my actual disagreement, which I promise is coming — the original critique isn’t really about AI per se. It’s about the communicative gesture. It’s about what you’re signaling to the other person when you respond to their human question with a machine-generated wall. And I think your framing of “it doesn’t hold for AI text that is checked and edited” actually concedes more than you intend, because now we’re having a debate about degree of human involvement, which is a sliding scale with no obvious threshold, and sliding scales without obvious thresholds are, historically speaking, not a great foundation for the kind of clean “this is fine, this is not fine” distinction your argument seems to require.
Now — and I really am getting to my main point here, bear with me — there’s a secondary issue around what you call the “promises of the technology,” and I find this framing genuinely fascinating in a way I’d love to explore more, because it implies that we have some obligation to use AI in conversations in order to honor its potential, which is a strange kind of technological determinism that I’m not sure you fully intend but which I think is worth naming. Do we owe it to the technology to deploy it? I’m not convinced. Tools don’t have promises in the morally obligating sense. A hammer doesn’t feel let down when you use a screwdriver instead.
Anyway. My disagreement, in short — and I know this has been a long way around to get here, but I genuinely believe the context matters — is that the “it was edited by a human” defense, while not entirely without merit in narrow technical circumstances, functions in practice as an all-purpose absolution that people apply far too broadly, and the result is exactly what the original post was describing: a conversation that is technically answered but humanly abandoned.
But I’d be curious to hear what you think, genuinely.