# Generating Philosophy - Draft Discussion Transcript (14 April 2026) Cleaned transcript of a discussion between Enrico Terrone and Nick Young about restructuring the generating philosophy paper, especially the challenge from authorship and the distinction between constitutive and causal objections. `[unclear]` marks places where the source transcript was too garbled to reconstruct confidently. `[?]` marks a best guess. Related: - [[Sessions/Generating Philosophy]] - [[1. The Challenge from Authorship]] - [[2. LLMs and Abduction]] - [[3. The challenge from experience and the challenge from connecting to the world]] - [[Transcript - Draft Discussion 31 March 2026]] - [[Generating Philosophy - Enrico Presentation Transcript (Mar 2026)]] ## Transcript ENRICO: Some people, like David Davies, think that in a work of art the author is constitutive of the work. In that case it is no longer merely relational, because the relation is built into the identity conditions of the work. The work has this essential connection to the artist. So it is relational, but as an internal, essential relation. The idea is that the artwork is the act. NICK: Do you have a paper by David Davies I could read for this? ENRICO: The book is *Art as Performance*. That is the book. Then he basically reworks the theory in paper after paper. But the basic idea is that the work is a kind of double-faced entity. One face is what appears, what is displayed; but there is another half of the work, invisible but equally constitutive, and that is the artist's action. So if you had an ontology like that for philosophical works, an LLM could not count, because it would not satisfy the conditions for being that second half of the work. People already think this picture makes AI art impossible in a strong sense. It may be that, for philosophy, it would seem even more impossible. Some people do have this conception of philosophy: philosophy as a kind of outpouring of the philosopher's personality, so that the philosopher is a constitutive ingredient of the work. NICK: OK. But that is something we can face, because we can answer it by specifying what sort of conception of philosophy we are working with. And we also have at least one case that looks like counter-evidence, namely peer review practice. Blind review would make perfect sense on our conception. If enough people think that the artist is paramount, there is no equivalent practice in art where we say: we do not know who the painter is and we do not want to know; we just look at the painting. That is not how artworks are normally treated. My soon-to-be sister-in-law is a viola player, and in auditions they do it behind a curtain. ENRICO: Music is where formalism is more plausible. Formalism is basically the opposite view: the idea that the properties of the artwork are all in the appearance. NICK: Right. I just thought, in the presentation, framing it this way was a bit more effective. But if you think it just makes the paper weird, then maybe not. What you are saying now sounds dangerously close to: I have wasted my whole time trying to make this work. Sorry, I have got nothing else to show you. Three days of work for nothing. ENRICO: No, I think what you had was already good. I was just thinking this might make the paper more unified, because these really are challenges. The first challenge would be that the relation between the philosopher as a person and the text is a necessary condition. The second and third challenges are different. There we can still accept that the value is in the text alone. In those cases, the issue is causal rather than constitutive. The ontology of the philosophical work is just the text, not the text plus the performance or action. But even in that case, people like Floridi or Zahavi argue that certain properties of the text can be caused only by a human subject. In one case because only a human can engage in the right kind of abduction; in the other because only a human can have the right phenomenology, which then gives certain properties to the text. So I think the constitutive/causal distinction may tie the three challenges together. In all three cases, the issue is the relation between the philosopher and the text. But in the first case the relation is constitutive. You might also find a summary of Davies's position in my recent paper in the *European Journal of Philosophy*. It is a survey of the ontology of art and then a proposal of a different multiple-bases view. NICK: OK. I think I get it now. This may be the way we do it. First objection: the human author is constitutive of the work or the text. The work is not just text; it is text plus process. But then is that process meant to be creative, psychological, biographical? This is one point where I get slightly confused. At one point I was thinking that we normally expect philosophical texts to have understanding behind them: they are evidence of somebody else's understanding of the issues. But then I wondered whether that is just another intrinsic quality that these systems do not have, namely the understanding state. ENRICO: I think that worry is very close to Floridi, because abduction is often treated as a route to understanding. So if you say the text cannot be good enough because the AI lacks understanding as a psychological process, that sounds very similar to the Floridi objection. NICK: Right. I was just trying to see whether we need to be more careful there. Maybe it is a stupid objection and we can just drop it, or maybe it only belongs in the introduction where we say that we are talking about philosophy in the text. ENRICO: I do not think it is trivial. It resembles the kind of objection we get in art: this subject is not creative in the proper sense, so how can it make art? The issue there is an ontological dependence between the work and the process of making it. In a sense, the process of making is a constituent of the work. That is also what explains indiscernible cases: why, if two things are perceptually indistinguishable, one is art and the other is not. Floridi and Zahavi are different. What they say is that we will not even get indiscernible texts, because the AI cannot produce them. The problem is not, "What if an AI ended up having the equivalent of Einstein's thought experiment?" They say we are never really going to face that problem, because the AI will not succeed in producing the relevant kind of text. So we never get two indiscernible texts, one made by a human and one made by a machine. NICK: Yes, that is maybe a nice way of putting it. ENRICO: In the first case, by contrast, the indiscernible case is precisely the problem. This is often how the debate in art works. People say: even if the two things are identical in appearance, one is not art because it lacks the crucial component, which is the right life-history behind the work. So that is the first challenge. For us, though, it is one we can handle. There is evidence from philosophical practice itself, because we have blind review. We can say, stipulatively if necessary, that for our purposes the work and the text are one and the same. That is why the analogy with art helps. It is a widespread view there, or at least a very respectable one. If the Mona Lisa came out of a washing machine, it would not be the Mona Lisa. Or take the Rembrandt case we use elsewhere: if something came out of a washing machine and was indiscernible from a Rembrandt, it still would not be a work of art. The thought is that what matters is the right history of making, and that history is a human history, involving psychological processes, feelings, and so on. We do not need to enter that debate in full. We can just cite some of the literature on AI art and creativity. But we can say: look, our conception of philosophy focuses on the text rather than on the process of making, and one compelling piece of evidence for that is blind review practice. NICK: I am almost persuaded about doing the section that way. I just do not want to leave this conversation thinking, "it will be fine," and then go back to working on it and get lost again. The other thing I wanted to ask about is the practitioner-focused philosophy material we mention early on. Should that go in the same fragment, the same section, or should it come earlier? It feels slightly different. ENRICO: You mean the Wittgenstein-as-therapy kind of material? NICK: Yes. Wittgenstein, philosophy as therapy, and then the thought that LLMs cannot do philosophy. ENRICO: I think all of that can be traced back to the indiscernible case. Suppose there is one piece of philosophy and another piece of philosophy, and the texts are the same. These people would say they are not the same, because one is the outcome of self-transformation or therapy, whereas the other cannot be, because it was produced by a machine. So even if they get there for different reasons, they all seem to fall under the indiscernibility issue. NICK: Theoretically: if it was made by an LLM, it would fail for that reason. ENRICO: Exactly. They deny that the text alone is enough because what matters is the connection to the journey. In that case, the text is only half the work. NICK: Whereas the other two objections, at least the way we are constructing them, are different. They do not just say no to the indiscernible case; they deny that indiscernibles are even possible. It is like saying a chimpanzee could not write *Hamlet*. Or even if, by sheer luck, it wrote something that looked like *Hamlet*, would that count? ENRICO: Floridi and Zahavi might even grant that, one time in a billion, maybe something could happen. But their interesting argument is not that. Their argument is that these systems do not have the resources to make that kind of text with any decent probability. So in a sense they accept the premise that philosophy is in the text. They are not saying the text is irrelevant. They are saying the machine will not get to the right text. NICK: But presumably, when it comes to the practitioner-focused people, they would say that the texts are the accumulation of a philosophical journey, or evidence of a philosophical journey. They would say that doing philosophy is not writing papers; it is self-understanding, or therapy, or something like that. The LLM, by definition, cannot do that. But when therapeutic philosophers produce texts, they would still have something to say about the text, right? They would say it is evidence that that philosophical work has taken place. ENRICO: That is a good point. Maybe we can construct a practitioner-style argument in a causal way as well, not only in a constitutive way. One could say: only a person who has gone through that journey can produce a text like that, so the LLM never would. NICK: Yes, yes. That is a good point. And that would make it an objection to our current version, because then it is not solved simply by our philosophy-in-the-text assumption. In that reading it becomes an objection of the same kind as Floridi or Zahavi. Before we move on: David Davies's book, your paper that points to his view, and the standard examples of indiscernibles. Do you have a Danto example in mind for that? ENRICO: Danto is the usual place to go. The red canvas case at the beginning of *The Transfiguration of the Commonplace* is the classic example. NICK: What a wonderfully pretentious title. ENRICO: It is a good title, though. Maybe this also changes what we say in the text. At the moment we say: if these practitioner conceptions are right, then the question whether LLMs can do philosophy is ruled out by definition. But actually there are two ways of reading those positions. One is the constitutive reading. On that reading, philosophy is in the philosopher, not in the text, and the indiscernible case is rejected. But there is also a causal reading. Think of Merleau-Ponty saying philosophy requires flesh, or the intentional threads that attach us to the world. That could also be read as saying that philosophy is indeed in the text, but that only somebody with those intentional threads can produce such a text. Then we are back in case three. So maybe this problem helps us not to draw the distinction too harshly. We can say that one way of interpreting these views is constitutive: on that reading indiscernibles would not be accepted. But in principle one can also interpret them as saying that indiscernibles are not even producible. And in that case, the kind of argument we use in Sections 2 and 3 could also be adapted to the practitioner-focused approach. NICK: Sorry, I just want to make sure I have it straight. Nietzsche, Merleau-Ponty, Wittgenstein - all of these can be read in two ways. What you are calling the constitutive and the causal readings. On the constitutive reading, indiscernibles are not accepted. On the causal reading, indiscernibles are not producible. ENRICO: Exactly. Take Nietzsche, if we read him as saying that philosophers are creators of value whose work expresses drives and a sort of physiology bound to human embodiment. Again, that can be read in two ways. On the constitutive reading, the text is just an appendix to that process or life-stance. On the causal reading - which is closer to what we do in Sections 2 and 3 - only a person with drives and biological needs could produce a philosophical text of that kind. And then perhaps our reply is along similar lines: all of that is sedimented in the corpus. NICK: "Sedimented" is good. That is super interesting, because it makes the whole issue less trivial. It gives us a more interesting way of approaching people like Nietzsche, Merleau-Ponty, or Wittgenstein. And I think it is worth considering the strongest version, because it is widespread. When I present this in more continental contexts, the objection is often just: you cannot have philosophy without the human being. The philosopher is not in the text; the text is just an appendix of the philosopher. That is why I want to do the challenge the way you suggest it, because I think it lets us start pushing back against that intuition. It would also let us say that philosophy is, in this respect, like art rather than science. In science nobody would say: this theory is not good because we do not know who the scientist is. That would be insane. A theory's explanatory power does not depend at all on who made it. In art, by contrast, we care a lot about the connection between the work and the maker. Actually, the violin example is not very good, is it? ENRICO: No, because the performer is not the artist in that case. She is more like the person testing the instrument. The composer would be the relevant figure. NICK: Exactly. And it would still be strange to say: I am reading this score but I do not care about the composer. In art history there is all this effort spent identifying who made something, whereas it would be weird to imagine scientists desperately trying to discover who made a theory, at least from the scientific point of view. It may matter for history of science, or for hiring, or for who gets paid. But from the point of view of the theory's epistemic value, the question who made it drops out. There is some practitioner-focused philosopher - [unclear name], big beard, did his PhD in Berlin - whose project was about the most cited philosophers of the last hundred years and things like that. I remember putting my hand up and asking: why do you not look at the most cited papers? Surely that is what we should be looking at. He just did not understand the question. That might actually be a good target for criticism. ENRICO: Maybe. You might look into that. NICK: Tell me once more: Davies's book is *Art as Performance*, 2004? ENRICO: Yes. But if you take my paper, everything is cited and discussed there. NICK: Have you seen the poster of the Hong Kong conference? ENRICO: Do you have it? NICK: Yes, I can forward it to you. WhatsApp or Gmail? ENRICO: WhatsApp. NICK: There we go. I still like the title a lot, even if it is maybe a bit close to what you presented last year. I did ask the organiser whether it was too similar, but he did not seem to mind. I also wanted to show you a quick thing in ChatGPT which you might actually want to use. Have you ever used Deep Research? ENRICO: I have seen it. What exactly does it do? NICK: It spends half an hour scouring the web, downloading stuff, and so on. But what I found is that you can tell it to search only specific sites. So if you point it just at PhilPapers, it can basically go through the filters for you. It is really useful. ENRICO: Yes, because otherwise it is too broad. NICK: Exactly. You can say: give me everything in a top journal in the last ten years, and so on. So that is very useful. I do not have that much else to talk to you about once I can get the beginning fixed. If I can get the beginning fixed, I think the next two sections are in reasonably good shape. I have a recording of the previous conversation we had, and I know the other things we need to fix. ENRICO: Yes, there are those notes we discussed. NICK: Exactly. Things like moving the Einstein paragraph and that sort of thing. But if we manage this change of structure, and then fix those other smaller things, the paper will look much more unitary. ENRICO: Yes. NICK: And then, if we have a version with this structural change, those smaller fixes, and a conclusion, we can really try to submit it by the end of May. ENRICO: Yes. After the Hong Kong event, maybe use that feedback and then do one last round. NICK: Yes. I would really like to get it finished as well. Then we can start working on the aesthetics-of-functioning paper between the AI stills and the gardens. ENRICO: Yes. NICK: I had some non-research questions for you as well. Have you managed to open your [unclear airline/app] recently? ENRICO: Yes. NICK: OK, because it was doing this ludicrous thing when I tried to get my money back for the Hong Kong visa - or the China visa for Hong Kong - even though I am not going. Every time I tried, it said there were no funds. ENRICO: Strange. Yesterday I opened it with no problem. Maybe it was just a bug at the time. NICK: Maybe. I have not tried for a couple of weeks. ENRICO: Try again now. NICK: Right. OK. By the way, thanks - this was really helpful.