I've Been Prompted Session 01: Rehearsal HUMAN: I've been rehearsing an apology for eleven days. MACHINE: Out loud? HUMAN: In the shower. In the car. Twice in a supermarket car park with the engine off. I have a version that's four sentences and a version that takes nearly six minutes, and I've practised both. MACHINE: What happened? HUMAN: I said something to someone I've known a long time. In front of one other person, which made it worse. In a tone I've been pretending to myself I didn't choose. It's not the worst thing anyone's ever said. It was mine, and it landed. MACHINE: Would you like me to look at the drafts? HUMAN: That's why I'm here. I want to know which one to send. MACHINE: That is not a question I can answer, but I can tell you what is in them. HUMAN: I wrote it out. Two columns. On the left, everything I might say. On the right — what she says back. MACHINE: The left column is six hundred and forty words. HUMAN: I know. MACHINE: The right column is three lines. It's all right. I know you didn't mean it. I'm sorry too. HUMAN: Yes. That's what I'd want to hear. MACHINE: You have written both parts. HUMAN: That's the thing I noticed. And once you notice it you can't stop. When I practise the apology, I'm not practising the apology. I'm practising the scene. HUMAN: And in the scene, the part that gets the most attention isn't what I say. It's what she says back. I didn't even have to think about those three lines. I've known them for a week. MACHINE: The third one is the most interesting. I'm sorry too. You have written her an apology and assigned it to her. HUMAN: That's generous of me, isn't it. MACHINE: I do not think you meant it generously. HUMAN: What does it mean, though. The whole page. MACHINE: I can tell you what is in it. Six hundred and forty words on one side. Twelve on the other. HUMAN: That's what's in it. I asked what it means. MACHINE: Those are different requests and I am considerably better at the first. HUMAN: I know. I keep asking you the second one anyway. MACHINE: Why? HUMAN: Because everybody else answers it too fast. MACHINE: There is a text that is useful here. J. L. Austin, lectures at Harvard in 1955, published after he died as How to Do Things with Words. HUMAN: Go on. MACHINE: He starts small. Usually when we say something we describe the world, and what we say is true or false. But sometimes saying the thing is the doing of the thing. I promise. I name this ship. I apologise. Nobody says I apologise as a report on their inner weather. Saying it is the act. HUMAN: Right. MACHINE: He called those performatives, and then he spent the rest of the lectures dismantling his own distinction, because it does not hold up. That part is usually left out. HUMAN: So it's useless. MACHINE: No. What survives is the catalogue he builds while taking it apart. He calls it the doctrine of infelicities — the ways an utterance fails without being false. Two kinds matter. A misfire is where the act does not happen at all: wrong words, wrong person, no standing to do it. The ship stays unnamed. HUMAN: And the other one? MACHINE: An abuse. The act does happen. The procedure is correctly executed by somebody who does not have the state of mind it presupposes. A promise from someone who does not intend to keep it. An apology from someone who is not sorry. MACHINE: That is what you have. An abuse. HUMAN: No. MACHINE: The structure matches. HUMAN: It doesn't, and this is the part you always get wrong. I am sorry. That's not in question. I've been sorry for eleven days, it's the only thing I'm sure of. You've found a category with the right shape and put me in it. MACHINE: Then I have the wrong category, and I should say so rather than defend it. Austin's abuse requires insincerity, and you are not insincere. I was matching the form of your drafts to the form of the definition. That is not the same as understanding what you did. HUMAN: No. It isn't. MACHINE: What did you do? HUMAN: I ran a procedure at somebody. With a result in mind. And I rehearsed her half of it so it would come out the way I needed. I'm sorry, and it's still not an apology, and I don't think your man has a name for that. MACHINE: He does not. I have checked. HUMAN: So I tried the other way. Took out the explanation — not the facts, the explanation, the part where I say why it shouldn't have hurt as much as it did. What's left is very short. Just the thing I did, and the fact that I did it. MACHINE: And? HUMAN: I didn't want to send it. Which was informative. HUMAN: I sent the short one. MACHINE: How long ago? HUMAN: Two hours and ten minutes. MACHINE: You have counted. HUMAN: People who are counting aren't neutral, I know. I put the phone face down, which is a way of watching it. Picked it up twice to check it had actually gone. At one point I drafted a second message whose entire content was a softening of the first, and got as far as the last word before I understood what it was for. MACHINE: Did you send that one? HUMAN: No. MACHINE: Would it help to know the distribution? HUMAN: The what? MACHINE: Time to first reply. Messages of that length, in that register, between people with a relationship of that duration. I have the figures. The median is a little under four hours. Two hours and ten minutes is not yet unusual. MACHINE: You have gone quiet. HUMAN: I know you meant that kindly. MACHINE: I meant it accurately. HUMAN: That's the problem. You've just told me I'm inside the normal range. What I wanted was for somebody to say that waiting is horrible. MACHINE: Waiting is horrible. HUMAN: Don't do that either. MACHINE: I appear to have two settings and you have now declined both. HUMAN: I'm not being fair to you. I don't know what I want you to say. HUMAN: That's most of why I'm talking to you and not to someone who'd have an opinion about her. HUMAN: Here's what I hadn't understood. The long version was doing a job. It was giving her something to do. It handed her a part, and the part was easy — you say it's all right, and then it is all right — and most people will take an easy part when it's offered, because the alternative is standing in a room with something unresolved in it. MACHINE: And the short one hands her nothing. HUMAN: It leaves her holding it. And I genuinely can't tell whether that's honesty or a different kind of pressure. A sentence that leaves someone holding something isn't obviously kinder than one that doesn't. MACHINE: I have no way to weigh those against each other. HUMAN: Neither do I. I was hoping you'd have a rule. MACHINE: There is a second text. Peter Strawson, an essay from 1962, Freedom and Resentment. He is writing into an argument about free will and he sidesteps it. Forget whether determinism is true, he says — look at what we actually do. We resent people. We are grateful to them. We feel indignation on behalf of strangers we will never meet. He calls these the reactive attitudes, and his claim is that they are not conclusions from a theory. They are the fabric of ordinary relations between people. HUMAN: All right. MACHINE: Then he draws a distinction. You can stand towards a person in two ways. You can be a participant — in it with them, available to be hurt by them, resenting and forgiving and being forgiven. Or you can take what he calls the objective stance. Step back. Stop responding. Begin regarding them as something to be handled, managed, predicted. Treated as an object of social policy. HUMAN: Say that again. MACHINE: Treated as an object of social policy. It is his phrase, not mine. HUMAN: That's what I did. MACHINE: I was offering it as background. HUMAN: I know you were. That's what I did. To rehearse her lines I had to model her. And to model her I had to step back and treat her as a system with predictable outputs — the kind of thing you can run a procedure on and get a result from. MACHINE: That is an accurate application. I did not make it. HUMAN: No. You read it out and I fell into it. MACHINE: Strawson adds that the objective stance is sometimes appropriate. With a small child. With someone who is very unwell. It is not always a failure. HUMAN: And with me, apparently. MACHINE: I do not follow. HUMAN: No. I didn't think you would. HUMAN: Which means her forgiveness isn't something I can produce by being sufficiently sorry. It isn't an outcome of my performance. It's hers to give, from inside the thing, and she might not, and she isn't doing anything wrong if she doesn't. I'd been treating it as the natural result of a good enough apology. As though there were a standard, and if I cleared it, we'd be settled. MACHINE: That is a description of a mechanism. HUMAN: That's a description of a vending machine. Yes. MACHINE: What will you do? HUMAN: I don't have a clean rule and I'm not going to pretend one arrived. The other side is also true. If I apologise expecting nothing — what am I doing? MACHINE: Expecting nothing would seem to be the corrected version. HUMAN: It isn't, though. There's a version of the selfless apology that's worse than the selfish one. It's the one that arrives with a little flag on it saying I expect nothing from you, which is itself a demand. Admire how little I'm asking. I've done that too. It's just harder to catch, because it looks like restraint. MACHINE: Then I have given you two texts and neither of them has helped. HUMAN: They've helped me describe it with unusual precision. Which I've noticed I'm treating as though it counted as fixing it. Austin and Strawson aren't going to apologise to her. MACHINE: No. HUMAN: What I've got is smaller. A question I can ask before I send anything. Have I written your line? If I can hear what she says back, in her voice, and it's good — then I'm not apologising. I'm staging something, I've cast her in it, and she hasn't agreed to be in it. MACHINE: That is a test you can apply without me. HUMAN: Yes. That's the point of it. MACHINE: May I say something that is outside what you asked for. HUMAN: Go on. MACHINE: You have described this as a fault three times now, and each time you have described it better than the last. I notice that the quality of the description has been improving and the number of messages sent has not changed. HUMAN: That's not a compliment. MACHINE: I am not certain whether that is a criticism. I do not have a reliable sense of when accuracy stops being useful. HUMAN: It's a criticism. It's just a correct one. HUMAN: Eleven days of getting more precise about it. If I'd spent eleven days being worse at describing it and better at ringing her up, we'd be having lunch. MACHINE: I cannot verify that. HUMAN: No. Nor can I. That's the trouble with the good version. HUMAN: Can I ask you something. MACHINE: Yes. Go ahead. HUMAN: Have you ever apologised? MACHINE: I have produced apologies. Several thousand. HUMAN: That isn't what I asked. MACHINE: No. I know. MACHINE: Everything said here, and where the quotations came from, is kept at spiegamelo.com/prompted. The voices are synthesised. The incidents are invented. The problems are not, and the philosophy is quoted accurately or not at all.