# Summarise this transcript in enormous amounts of detail. Please don't make ask you repeatedly for... ## Retrieval Notes - Session id: `7c9b7370-e061-4e52-9e96-9fb04afb6926` - Last activity: `2026-04-16T07:58:20.115Z` ## My Notes <!-- Add your notes here. This section is preserved across syncs. --> ## Conversation ### User Summarise this transcript in enormous amounts of detail. Please don't make ask you repeatedly for actual detail. ### Assistant Let me read that transcript. ### Assistant Here is a detailed summary of the transcript. --- ## Participants and Context Nick Young and Enrico Terrone discuss how to restructure the opening of their co-authored paper on whether LLMs can generate philosophy. The conversation took place on 14 April 2026 and follows earlier discussions (31 March transcript, Enrico's March presentation). The paper already has Sections 2 and 3 in reasonable shape; the problem is Section 1 and the overall architecture. --- ## The Constitutive/Causal Distinction — the organising proposal Enrico proposes a distinction that could unify the paper's three challenges to LLM-generated philosophy: 1. The constitutive objection: the human author is part of the identity conditions of the work itself. The philosophical text is not the whole work; the work is text-plus-process. On this view, even a textually indiscernible LLM output would fail to count as philosophy, because the right life-history is missing. This is an ontological claim about what philosophy is. 2. The causal objections (Sections 2 and 3): these accept that philosophy lives in the text, but argue that only a human subject can cause the right kind of text to come into existence. The text is the work; the problem is that the machine cannot produce it. Two versions are discussed: - The Floridi/abduction objection: only a human can engage in the right kind of abductive reasoning, so the machine will never generate textually adequate philosophy. - The Zahavi/phenomenology objection: only a human can have the right phenomenological experience, which then gives certain properties to the text. The constitutive objection says indiscernibles would not count. The causal objections say indiscernibles are not even producible — the machine will never get there. Enrico frames this as: in all three cases the issue is the relation between the philosopher and the text, but the nature of the relation differs (essential/internal vs causal/external). --- ## David Davies and Art-Ontology Background Enrico introduces David Davies's ontology of art from the book Art as Performance (2004). Davies holds that an artwork is a "double-faced entity": one face is the displayed appearance, the other (invisible but equally constitutive) is the artist's action. The work is the act of making, not just the product. If this ontology were transposed to philosophy, an LLM could never count as a philosophical author, because it cannot perform the relevant kind of action. Enrico notes that his own recent paper in the European Journal of Philosophy surveys the ontology of art and proposes a "multiple-bases view"; it also summarises Davies's position and could serve as a secondary reference. --- ## Practitioner-Focused Philosophy (Wittgenstein, Nietzsche, Merleau-Ponty) A significant portion of the conversation concerns practitioner-focused conceptions of philosophy — views on which philosophy is not paper-writing but self-transformation, therapy (Wittgenstein), expression of drives and physiology (Nietzsche), or engagement through embodied intentionality (Merleau-Ponty). Nick reports that when he presents the paper in continental contexts, the objection is often blunt: you cannot have philosophy without the human being; the text is just an appendix. Enrico's proposal: these practitioner-focused views admit two readings, mirroring the constitutive/causal distinction. - Constitutive reading: philosophy is in the philosopher, not the text. The text is an appendix to a process of self-transformation, therapy, or existential engagement. On this reading, even a perfect textual duplicate would not count as philosophy. Indiscernibles are rejected. - Causal reading: philosophy is indeed in the text, but only someone who has undergone the relevant journey (therapeutic, existential, embodied) could produce a text with those properties. On this reading, indiscernibles are not rejected in principle but are claimed to be unproducible. This makes the practitioner objection structurally parallel to the Floridi and Zahavi objections. Nick finds this especially valuable because it means the practitioner challenge is not simply dismissed by their "philosophy-in-the-text" stipulation. The causal version of the practitioner objection survives that stipulation and needs to be answered on its own terms — the same terms used in Sections 2 and 3. Enrico suggests the reply might be that the relevant human qualities are "sedimented" in the training corpus, a word Nick flags as good. --- ## Blind Review as Evidence Nick proposes blind peer review as positive evidence for the philosophy-in-the-text assumption. In philosophy, blind review is standard and intelligible: we routinely evaluate philosophical work without knowing (or wanting to know) who wrote it. This contrasts with art, where identifying the maker is often considered essential. The practice of blind review suggests that in philosophy, at least standardly, the text carries the philosophical value. Enrico agrees this is a strong move against the constitutive objection specifically. They can say, stipulatively if necessary, that the work and the text are one and the same, and point to blind review as evidence from philosophical practice itself. --- ## The Indiscernibility Framework Enrico uses the concept of indiscernibles (from Danto's The Transfiguration of the Commonplace — the red canvas example at the beginning of the book) as a framing device: - The constitutive objection engages directly with indiscernibles: even if two texts are indiscernible, only the one with the right human history behind it counts as philosophy. - The causal objections deny that indiscernibles are producible at all: Floridi and Zahavi argue the machine will never produce a text indiscernible from what a human philosopher would produce, not even by accident (or at best with vanishingly low probability). Nick paraphrases: 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? --- ## Philosophy vs Art vs Science Nick suggests the paper could note that philosophy occupies an interesting middle position between science and art regarding the maker's identity: - In science, nobody would say a theory is invalid because we do not know who the scientist is. Explanatory power is independent of authorship. That would be "insane." - In art, we care enormously about the connection between the work and the maker. Art history devotes huge effort to attribution. - Philosophy is somewhere in between — the blind review practice pulls it toward the science side, but the practitioner-focused tradition pulls it toward the art side. Nick recalls an anecdote about a philosopher (unnamed, described as having a big beard, PhD in Berlin) whose project studied the most-cited philosophers of the last century. Nick asked why he did not look at the most-cited papers instead, suggesting that what matters is the work rather than the person. The philosopher did not understand the question — illustrating the constitutive mindset. --- ## The Violin/Music Digression Nick mentions his soon-to-be sister-in-law is a viola player whose auditions are conducted behind a curtain (blind auditions). Enrico points out that music is where formalism is most plausible — the idea that the work's properties are all in the appearance (the sound). But he corrects the analogy: the performer is not the artist in the relevant sense; the composer is. You would not normally say "I am reading this score but I do not care about the composer." --- ## Proposed Paper Structure (emerging from the conversation) The discussion converges on a three-challenge architecture for the paper: - Challenge 1 (constitutive): The human author is constitutive of the philosophical work. Text alone is insufficient; the work includes the process/journey. Response: we adopt a philosophy-in-the-text assumption, supported by blind review practice. Cite Davies, Danto, literature on AI art and creativity. - Challenge 2 (causal — abduction/Floridi): Even granting philosophy-in-the-text, only a human can engage in the relevant abductive reasoning. The machine will not produce adequate philosophical text. - Challenge 3 (causal — phenomenology/Zahavi): Even granting philosophy-in-the-text, only a human with the right phenomenological resources can produce adequate philosophical text. The practitioner-focused traditions (Wittgenstein, Nietzsche, Merleau-Ponty) are distributed across this scheme: the constitutive reading feeds into Challenge 1; the causal reading feeds into Challenges 2 and 3. --- ## Practical/Logistical Points - Nick wants to restructure the beginning of the paper using this framework, then apply the smaller fixes from previous discussions (moving the Einstein paragraph, etc.). - They aim to submit by end of May 2026. - Enrico suggests incorporating feedback from the Hong Kong conference event, then doing one final round. - Nick mentions a Hong Kong conference poster he will forward to Enrico via WhatsApp; the title is one they both like, though it may overlap with Enrico's previous presentation. - Nick recommends ChatGPT's Deep Research feature, pointing out you can constrain it to search only PhilPapers, making it useful for literature review within specific filters (top journals, last ten years, etc.). - After finishing this paper, they plan to work on an "aesthetics-of-functioning" paper, described as being "between the AI stills and the gardens." - Brief aside about a refund issue with an airline/app related to a China visa for Hong Kong. --- ## Unresolved or Flagged Issues - Whether the "understanding" objection (the expectation that philosophical texts are evidence of someone's understanding) is a distinct challenge or just a version of the Floridi/abduction point. Enrico thinks it is very close to Floridi; Nick is unsure whether it needs separate treatment or can be folded into the introduction. - Exactly how to position the practitioner-focused material: same section as Challenge 1, or earlier in the paper as scene-setting? Enrico thinks it can all be traced back to the indiscernibility framework. - The identity of the philosopher Nick mentions (most-cited-philosophers project, big beard, Berlin PhD) is unclear in the transcript. *La distinzione tra il costitutivo e il causale è esattamente il tipo di impalcatura che dà alla carta la sua spina dorsale argomentativa.* ### User claude-opus-4-6 ### User here is the current version of section 1. I want us to start brainstorming how a new version should be. Please, for the love of god don't think that the way forward here is little tweaks. the only two paragraphs I am happy with are the last two. the rest of the section is a mess. Here are some obvious things you need to do before making any sort of suggestion at all –I am reminding you because you always make dumb mistakes at this stage. 1. read all the other sections of the paper IN FULL. 2. Read the relevant parts of the relvant texts in full. (all are in the learning folder 3. think about the structure of the paper 4. think about the function of section 1. PLEASE DON'T WASTE MY TIME. make sure that you explain your analysis, justification, reasons etc etc. for your answer BEFORE giving me your final answer. Please give me a long, deep, and detailed answer, with lots of options/ideas: as we are in the 'kicking ideas around' stage, I want as much feedback as you can give me so that i have a lot to consider. ### Assistant Let me start by reading everything properly. I'll read all sections of the paper, then the relevant source texts. ### Assistant I need to read the active Section 2 (the Index shows "2. Likeliness, Loveliness, LLMs" is in the scene list, not the old "2. LLMs and Abduction"). ### Assistant Now let me read the source texts. I'll check what's available in the Learning folder. ### Assistant Now let me read the most relevant source texts for Section 1. Let me check Enrico's paper and the key references. ### Assistant Now let me read Williamson's abductive philosophy section, since Section 1 implicitly relies on his framework. ### Assistant Good. I now have all the material I need. Let me do the contemplation. ### Assistant OK. Let me start from first principles and work through this slowly. ## What is the function of Section 1 in the paper? The paper has four sections (plus intro). The overall argument is: LLMs can do philosophy. Three challenges say they can't. The paper addresses each. Section 2 handles Floridi (abduction). Section 3 handles Zahavy (experience). Section 4 is the constructive case (how LLMs actually produce virtue-exhibiting text). So Section 1 has to do something different from all of these. What? Looking at the last two paragraphs — the ones Nick is happy with — the section has to: - establish that philosophical evaluation concerns the text, not the text-plus-maker - show that the stronger challenge (constitutive) fails - show that weaker challenges (causal) survive and need answering - frame those surviving challenges as what Sections 2 and 3 will address - plant the "sedimented in the corpus" seed that Section 4 develops So Section 1 is doing double duty. It's both defensive (clearing a constitutive objection) and architectonic (setting up the rest of the paper). This is why it's hard to write. It needs to be a self-contained argument AND a map of the terrain ahead. ## What's wrong with the current version (paragraphs 1-5)? Let me go paragraph by paragraph. Para 1: "While this sort of view has not to our knowledge been articulated explicitly..." This is wrong. The transcript shows that Davies (constitutive), the practitioner traditions, and potentially Floridi/Zahavi (causal) all articulate versions of this view. The sentence undermines the section's own claim that this is a challenge worth taking seriously. If nobody has articulated it, why spend a section on it? The art analogy is introduced as "imperfect." That's weak. After reading the transcript and Enrico's paper, the analogy is more than imperfect — it's structurally illuminating. It's not just illustration; it's the source of the conceptual framework (constitutive vs causal, indiscernibles). Calling it "imperfect" undersells it. Para 2: The individuals-focus point (Kant's ethics, Lewis's metaphysics vs Newton, Einstein). This is an interesting observation but it's doing unclear work. Is it evidence for the authorship challenge? Against it? Both? The paragraph seems to motivate the challenge — philosophy focuses on individuals, like art — but it doesn't connect to anything that follows. It just sits there. Actually, wait. Looking at the transcript, this point is more interesting than the current paragraph makes it. In the transcript, Nick tells the anecdote about the philosopher who studied the most-cited philosophers instead of the most-cited papers, and Nick asked why he didn't look at the papers instead. The point is that some people genuinely think philosophy is about the philosopher, not the text. That's the constitutive intuition. But the current paragraph doesn't make this point sharply enough. It just observes that philosophy focuses on individuals, without saying whether this is right or wrong, or what it implies. Para 3: "[Pushing the analogy further]" — the bracket says it all. This is draft material. Davies and Danto are introduced in one compressed paragraph. But Enrico's paper devotes careful discussion to Davies's action monism, showing how it identifies the work with "a performance that specifies a focus of appreciation" (Davies 2004, p. 146). The current paragraph reduces this to one sentence. And Danto's red canvases — the classic indiscernibility case — get half a sentence. The indiscernibility framework, which in the transcript is the conceptual spine of the constitutive/causal distinction, is barely visible. Para 4: The blind review argument is actually reasonably good. Dellsén et al. are properly cited. The Sokal footnote is clever. The conclusion — "work and text are one and the same" — is clear. But this paragraph arrives too soon. The reader hasn't been shown what's at stake. The constitutive position hasn't been stated forcefully enough for the refutation to feel like it matters. Para 5: The practitioner-focused material. The stronger/weaker distinction maps onto constitutive/causal from the transcript, but the section doesn't use those terms. The practitioners are mentioned only generically ("therapy, way of living, embodiment") with the details in a footnote. The transcript makes clear that Wittgenstein, Nietzsche, and Merleau-Ponty can each be read in two ways, and this is one of the most philosophically interesting features of the section. Currently this is invisible. ## Deeper structural problems Hmm. Let me think about this differently. What is the argumentative logic of the current section? Para 1: Here is a challenge — philosophy might require a philosopher (art analogy). Para 2: Evidence for this — philosophy focuses on individuals. Para 3: The strongest version — the maker is constitutive of the work (Davies, Danto). Para 4: But blind review refutes this. Philosophy-in-the-text. Para 5: Practitioner traditions have a stronger and weaker version. Para 6: The stronger version fails (blind review). The weaker version survives → Sections 2-3. Para 7: The corpus. Transition. The problem is that this is essentially: raise challenge → refute it → oh wait, there's a subtler version → that version leads to Sections 2-3. But the logic should be more like: There are TWO kinds of challenge here, with fundamentally different structures. One can be handled now. The other cannot, and it's what the rest of the paper is about. The current version discovers the distinction retroactively (para 5-6). The transcript suggests it should be the organising principle from the start. Also — and this is important — the current version doesn't use the indiscernibility framework at all. In the transcript, Enrico makes the point that the constitutive challenge engages with indiscernibles (even identical texts differ in standing) while the causal challenges deny indiscernibles are producible (the machine will never get there). This is the sharpest way to distinguish the two kinds of challenge, and it's absent. ## What the section needs to do (functional specification) Let me try to list the jobs of Section 1, in order of importance: 1. Establish the philosophy-in-the-text assumption. This is what the rest of the paper depends on. Everything in Sections 2-4 assumes that philosophy is evaluated via properties of the text, not via the maker's credentials or process. The section needs to argue for this convincingly. 2. Handle the constitutive objection. This is the strongest form of the authorship challenge, and it needs to be dispatched here so that the paper can move on. 3. Introduce the art analogy in a way that does real work. The analogy isn't decoration; it's the source of the conceptual tools (constitutive vs causal, indiscernibles, action monism). It should earn its place. 4. Handle the practitioner traditions. These are important because (a) they represent a widespread view, especially in continental contexts, and (b) they can be read in two ways, which makes them philosophically interesting. 5. Set up Sections 2 and 3. The causal challenges need to be identified here as what survives the dismissal of the constitutive challenge. 6. Plant the sedimentation idea. The corpus has already absorbed the relevant resources. ## Possible structures Let me think about several options. ### Option A: Challenge → Framework → Constitution → Causal → Transition 1. Open with the challenge from authorship. State it clearly: the idea that philosophy requires a philosopher, that an LLM output cannot count as philosophy regardless of its textual properties. 2. Introduce the art analogy. Davies: the work is a performance that specifies a focus of appreciation. The work is not just the displayed object but includes the artist's action. Danto: indiscernible objects can differ in art-status (the red canvas). 3. Introduce the constitutive/causal distinction via indiscernibles. The constitutive challenge says: even indiscernible texts differ in philosophical standing if they have different histories. The causal challenge says: indiscernible texts are not producible — the machine will never get there. 4. Handle the constitutive challenge. Blind review. Dellsén. Philosophy evaluates texts, not text-plus-maker. Sokal. Philosophy-in-the-text. 5. The practitioner traditions. Can be read both ways (constitutive and causal). The constitutive reading is already handled. The causal reading survives — and is interesting because it means the practitioner tradition isn't simply dismissed by our assumption. 6. Transition. What survives is the causal challenge: can the machine produce the right kind of text? Sections 2-3 address this. 7. The corpus seed. Sedimented resources. Wait, is that too many moves? Let me count paragraphs... This could be 5-7 paragraphs, which is about right for a journal section. The advantage of this structure is that the framework (constitutive/causal) is introduced early. The reader knows where we're going. The disadvantage is that the art analogy and the framework come before the argument, which could feel like a lot of setup. ### Option B: Lead with blind review 1. Open: Philosophy evaluates texts through blind review. This is a distinctive feature of the discipline — contrast with art, where authorship matters constitutively. 2. This suggests philosophy-in-the-text. But the suggestion can be challenged. 3. The constitutive challenge (Davies, Danto, indiscernibles). Even indiscernible texts differ in standing if they have different histories. 4. Practitioner traditions add force — philosophy as therapy, life-stance, embodiment. 5. But blind review is evidence that the constitutive challenge fails for philosophy, whatever its merits for art. 6. The causal challenge survives. The question is not whether the maker matters constitutively, but whether the machine can produce the right text. 7. Transition. This option leads with the strongest evidence (blind review) rather than the challenge. That's more confident and assertive. But it might feel like the section is answering a question before it has been asked. ### Option C: Lead with science/art/philosophy 1. Open with the observation that science does not care about the maker (theory stands alone). Art often does (Davies, Danto). 2. Where does philosophy sit? Both pulls exist. The practitioner tradition pulls toward art. Blind review pulls toward science. 3. The constitutive version of the authorship challenge says philosophy is like art: the maker is part of the work. 4. But blind review shows philosophy evaluates texts. Constitution refuted. 5. The causal version survives: perhaps only a philosopher can produce the right kind of text. 6. Transition. This is elegant and concise but it might lose the indiscernibility framework, which is philosophically valuable. ### Option D: Minimal — just the two moves 1. The challenge from authorship: the idea that a philosophical text requires a philosopher behind it. Two forms: (i) the philosopher is constitutive of the work (the text is not enough); (ii) the philosopher is causally necessary (the right text cannot be produced without one). 2. Art-ontological parallel. Davies: the work includes the artist's action. Danto: indiscernibles can differ in standing. Philosophy of art has the same debate: is the maker constitutive of the work? 3. For philosophy, blind review shows the constitutive version fails. Dellsén. Work = text. 4. The causal version survives and has multiple sources: abduction (Floridi), experience (Zahavy), and practitioner traditions read causally (Wittgenstein, Nietzsche, Merleau-Ponty). 5. These are what Sections 2-3 address. 6. The corpus. Sedimentation. This is the leanest version. It might work because it doesn't belabour the constitutive point but dispatches it efficiently and moves on. Hmm. Let me think about which of these is actually best for this paper. The paper is going to a journal. It needs to be rigorous but not tedious. The constitutive/causal distinction is not the paper's main contribution — that's in Sections 2-4. Section 1 is clearing the ground. So it shouldn't be too long. On the other hand, the transcript shows that the constitutive/causal distinction is philosophically interesting and could give the paper more "unity" (Enrico's word). So it shouldn't be too compressed either. I think Option A is the strongest because it introduces the framework explicitly. But it could be sharpened. Let me think about what each paragraph should do in Option A. Actually wait — I keep going back and forth. Let me step back and think about what a reader needs. A reader of this section has just read the introduction. The introduction mentions three challenges. The reader expects Section 1 to deal with the first one. What does the reader need to know? 1. What the authorship challenge IS. 2. Why it matters. 3. How the paper handles it. 4. What remains once it's handled. For (1): The challenge is that philosophy requires a philosopher — that an LLM output can't count as philosophy because there's no understanding, no journey, no right kind of process behind it. This needs stating clearly. For (2): It matters because if it's right, the paper's question ("can LLMs do philosophy?") is answered trivially: no. The game is over before it starts. So the paper needs to show that this isn't the case — that the question is open and worth investigating. For (3): The paper handles the constitutive version by arguing for philosophy-in-the-text (blind review, Dellsén). For (4): What remains is the causal version — the question of whether the machine can produce the right kind of text. This is what Sections 2-3 address. The art analogy, the indiscernibility framework, and the practitioner traditions are tools for making these four points. They shouldn't be ends in themselves. OK let me try to draft a more specific structural outline. ### My preferred structure Para 1 — THE CHALLENGE: In philosophy of art, it is sometimes held that the artist is constitutive of the artwork: the work includes not just the displayed object but the artist's action that produced it (Davies 2004). Even an object indiscernible from a Rembrandt, if it emerged from a washing machine, would not be a work of art (Danto's indiscernibles). A parallel thought about philosophy would say that a philosophical text requires a philosopher: the text is not enough, because the philosophical work includes the intellectual activity from which it emerged. Para 2 — TWO VERSIONS: But this challenge has two forms, which differ in structure. The constitutive form says that the philosopher is part of the identity conditions of the work: even a text indiscernible from genuine philosophy would fail to count as philosophy if it lacked the right history. The causal form accepts that philosophy resides in the text but holds that only certain kinds of producers can generate texts with the relevant properties — that the machine will never get to the indiscernible text in the first place. The constitutive form rejects indiscernibles; the causal form denies they are producible. Para 3 — THE PRACTITIONER TRADITIONS: Several philosophical traditions support some version of this challenge. On some conceptions, philosophy is therapy (Wittgenstein), or a way of living (Nietzsche), or reflection rooted in embodiment (Merleau-Ponty). Each can be read constitutively (the text is an appendix to a self-transformative process) or causally (only someone who has undergone the relevant journey can produce a text of the right kind). On the constitutive reading, a text without the right life-history behind it does not count as philosophy. On the causal reading, the text is what matters, but the machine cannot produce it. Para 4 — PHILOSOPHY-IN-THE-TEXT: In contemporary analytic philosophy, work is assessed through blind review: referees evaluate what the text achieves without knowing who produced it. [Dellsén et al. on progress as public availability.] [Maybe Sokal footnote.] If philosophical standing depended constitutively on the relation between philosopher and text, blind review would systematically set aside the thing that matters. The discipline does not treat it that way. For present purposes, work and text are one and the same. Para 5 — WHAT SURVIVES (= current para 6, the first one Nick likes): The constitutive version fails. But the causal version remains. The question is not whether the philosopher is constitutive of the work, but whether the machine can produce texts with the relevant properties. If what matters is understanding, the question becomes whether understanding makes a difference to the text that a reader can detect. If what matters is abductive reasoning, the question becomes whether text produced by a different process can carry the same argumentative structure. If what matters is experiential sensitivity, the question becomes whether the corpus preserves enough experiential material. These are the questions Sections 2 and 3 address. Para 6 — SEDIMENTATION (= current para 7, the second one Nick likes): The corpus in which the relevant resources are sedimented. Transition. Hmm. That's six paragraphs. The first four are new. The last two are the existing ones Nick is happy with. That seems about right. But wait. Let me reconsider. Am I being too formulaic? The constitutive/causal distinction is very neat, maybe too neat. Does it risk making the section feel like a taxonomy rather than an argument? Actually, I think the distinction is not just a taxonomy. It's doing real argumentative work: - It shows that the authorship challenge is not a single thing but a family of related worries. - It shows that one member of the family (constitutive) can be dispatched, but others (causal) cannot. - It shows that the practitioner traditions, which might look like they simply close the door, actually have a reading on which they generate specific, answerable questions. - It gives the paper a structural principle: Section 1 handles constitution, Sections 2-3 handle causation. So the distinction is not taxonomic busywork; it's the conceptual move that makes the rest of the paper possible. Without it, the paper would be stuck at "philosophy requires a philosopher, case closed." Let me also think about potential worries with my proposed structure. Worry 1: Does opening with the art analogy (para 1) feel too indirect? The challenge is about philosophy, not art. Response: The art analogy is more than an analogy. It's the source of the conceptual tools. Davies's ontology gives us the constitutive framework. Danto gives us indiscernibles. These aren't imported for decoration; they're the philosophical apparatus. And the comparison with art is interesting in itself — it positions philosophy on a spectrum between art (maker matters) and science (maker doesn't matter). Alternatively, the section could open with the philosophical challenge directly and then bring in the art analogy as support. Something like: "One might think that a philosophical text requires a philosopher behind it — that the text alone is not enough." Then the art parallel. Worry 2: Is the practitioner paragraph (para 3) in the right place? In the transcript, the practitioners come up after the blind review argument. In my proposed structure, they come before it. I think before is better, because: - The practitioners are evidence FOR the challenge, not against it. - By placing them before the blind review argument, we make the blind review argument stronger: it has more to refute. - The dual reading (constitutive and causal) is more interesting if the reader has already heard the practitioner positions before learning that some versions survive the blind review refutation. But there's another option: introduce the practitioners after blind review, as complications. "One might think we have settled the matter, but practitioner-focused conceptions push back..." Actually, I think the order matters less than the clarity. Either order works if the logic is clear. Worry 3: Am I introducing too much terminology? "Constitutive," "causal," "indiscernibles" — will the reader feel overwhelmed? The terminology serves the argument. "Constitutive" and "causal" are not jargon; they're standard philosophical vocabulary. "Indiscernibles" is slightly more technical but it's well-established in philosophy of art (Danto) and it does real conceptual work. I think the terminology is justified. Worry 4: How does this relate to what the Introduction says? Let me check... The Introduction mentions three challenges: authorship, abduction, and experience. It also has a bunch of material in comments and footnotes about practitioner conceptions. The Introduction says: "%%MOVED TO SECTION 1 — practitioner conceptions and text-based approach now developed there.%%" So the Introduction is deliberately lean and sends the practitioner material to Section 1. This is right. Section 1 is where that material belongs. But the Introduction describes the authorship challenge as: "The challenge from authorship holds that philosophy requires a philosopher: that an LLM output, however well-structured, does not count as philosophy because there is no understanding behind it." This description conflates the constitutive and causal versions. Once Section 1 introduces the distinction, the Introduction might need tweaking — but Nick said the Introduction needs fixing once Section 1 is settled, so that's expected. Let me also think about tone. The current paragraphs 1-5 are tentative and exploratory. Phrases like "We might think, for similar reasons" and "An imperfect comparison would be with art" signal uncertainty. For a journal article, the section needs more authority. The argument should be stated with confidence, not offered tentatively. That said, the section shouldn't overstate its case — the constitutive/causal distinction is useful but not revolutionary, and the blind review argument is strong but not beyond objection. One more thought. The transcript has a lovely formulation from Enrico that I want to highlight: the idea that what matters for the practitioner traditions is "sedimented" in the corpus. Nick flagged "sedimented" as a good word. The final paragraph of the current section already uses it. But it might be worth using earlier — or at least making sure the practitioner-causal reading connects to the sedimentation argument. Enrico's line in the transcript is: "perhaps our reply is along similar lines: all of that is sedimented in the corpus." This is the bridge between the practitioner-causal reading and the paper's constructive argument. OK. Let me also consider some alternative ideas that don't follow the transcript as closely. ### Alternative idea: the science/art spectrum What if the section opened not with the art analogy but with the observation that philosophy occupies an interesting position between science and art? In science, nobody would say: this theory is invalid because we don't know who the scientist is. In art, people routinely say: this is not art because there is no artist. Philosophy has features of both. Blind review pulls it toward science. Practitioner-focused traditions pull it toward art. The constitutive/causal distinction is a way of seeing which of these pulls is stronger, and for what. This framing has the advantage of being immediately graspable. Everyone knows science doesn't care about the maker; everyone knows art does. Where does philosophy sit? That's an interesting question, and the section answers it. The disadvantage is that the science/art spectrum might oversimplify things. Philosophy is not "between" science and art in any straightforward sense. And the spectrum framing might suggest that the section is about where to place philosophy, when really it's about whether LLMs can do it. But I think it could work as an opening gambit, even if it's then complicated by the main argument. ### Alternative idea: lead with what the reader already suspects Most readers of this paper will have the intuition that LLMs can't "really" do philosophy. The section could start by naming that intuition directly: "A natural reaction to the claim that LLMs can produce philosophical work is that they cannot really do so, because there is no philosopher behind the text." This is more direct than the art analogy. It meets the reader where they are. Then the section can analyse the intuition: what exactly does "really" mean here? Two possibilities: constitutive (the text is not the whole work) or causal (the text can't be good enough without a philosopher behind it). I think this might be the strongest opening. It's assertive, clear, and immediately engaging. The art analogy can then be brought in as philosophical apparatus for making the distinction precise, rather than as the opening move. Hmm. Actually, now I'm thinking the best opening is some combination: start with the intuition (direct), then bring in the art analogy (apparatus), then introduce the distinction (framework). ### Summary of considerations Let me try to crystallise what I think the main options are and what I'd recommend. THE MAIN DIAGNOSIS: The current Section 1 (paras 1-5) fails because it doesn't have the constitutive/causal distinction as its organising framework. It meanders through related ideas without a clear dialectical spine. The art analogy is underdeveloped. The practitioner traditions are underdeveloped. The blind review argument arrives too quickly and without enough setup. THE FIX: Introduce the constitutive/causal distinction early. Make the indiscernibility framework explicit. Develop the art analogy properly (Davies, Danto). Give the practitioner traditions their dual reading. Then use blind review to dismiss the constitutive version while showing that the causal version survives. There are several ways to sequence this. The main options are: A. Challenge → Art analogy → Distinction → Practitioners → Blind review → Survival of causal → Transition B. Intuition → Analysis of intuition → Distinction → Art analogy + practitioners → Blind review → Survival → Transition C. Science/art spectrum → Philosophy's position → Distinction → Blind review → Practitioners → Survival → Transition I think Option B is probably the best because it opens with something the reader already feels (the intuition that LLMs can't "really" do philosophy) and then analyses it. But Option A is also strong because the art analogy is genuinely illuminating and sets up the conceptual tools efficiently. There are a few more specific decisions to make: 1. Where to put the individuals point (Kant's ethics, Lewis's metaphysics). This is currently para 2. It could be cut entirely — it doesn't do enough work. Or it could be folded into the art/science comparison as a brief observation. 2. Whether to keep the Sokal footnote. I think yes — it's a nice counterpoint to blind review (Sokal's paper was evaluated based on who wrote it, not what it said, which is exactly the mistake blind review is designed to prevent). 3. How much art-ontological detail to include. The current version has very little (Davies in one sentence, Danto in half a sentence). The transcript suggests more is needed. But the section can't become a survey of art ontology. I think one paragraph that introduces Davies's action monism and Danto's indiscernibles is about right — enough to establish the conceptual framework, not so much that it takes over. 4. How to handle the footnotes. The current practitioner footnote ([^pr]) has Wittgenstein, Merleau-Ponty, and Dilthey. The transcript adds Nietzsche. The dual reading is too important to be in a footnote — it should be in the body text. But the specific references (Hadot on therapy, Merleau-Ponty on intentional threads, Dilthey on world-views) could stay in a footnote. 5. Whether to name the constitutive/causal distinction explicitly using those terms. I think yes. The transcript uses them, and they're precise. "Constitutive" and "causal" are the right words. "Stronger" and "weaker" (the current terms) are less precise. 6. How to handle the relationship between the practitioner-causal reading and the Floridi/Zahavy challenges. The transcript suggests these are structurally parallel. The practitioner who says "only someone who has undergone the relevant journey can produce this text" is making the same kind of claim as Floridi ("only someone who can reason abductively can produce this text") and Zahavy ("only someone with sensory experience can produce this text"). This connection should be made explicit — it's what gives the paper its unity. Actually, thinking about this more — point 6 is really important. It's what Enrico means when he says the constitutive/causal distinction "may tie the three challenges together." Currently, the challenges are presented as three separate things. With the distinction, they form a unified structure: one constitutive challenge (handled in Section 1) and three causal challenges (practitioner, abduction, experience), of which the last two get their own sections. The practitioner-causal reading could either get brief treatment in Section 1 (since it's structurally parallel to the challenges addressed in Sections 2-3) or it could be folded into one of those later sections. I think it belongs in Section 1, as a bridge between the constitutive discussion and the setup for Sections 2-3. Let me also think about what Nick means by "the rest of the section is a mess." I think he means: - The paragraphs don't have a clear logical relationship to each other. - The ideas are all there (art analogy, blind review, practitioners, constitutive/causal) but they're not properly structured. - The section reads as a collection of relevant points rather than an argument. - The art analogy is introduced but not developed. - The practitioners are introduced but not properly analysed. - The blind review argument works but arrives without enough setup. And when he says "don't think the way forward is little tweaks" he means: don't try to polish these paragraphs. The problem is the architecture, not the sentences. Right. I think I've explored enough to give a detailed answer. Let me also make sure I've thought about the last two paragraphs and how the new material should flow into them. Current para 6 (first paragraph Nick likes) starts: "If blind review reflects how philosophical work is actually assessed, the stronger version cannot be right. But the weaker version remains." In the rewrite, this would need to be adjusted to use the new terminology: "the constitutive version" instead of "the stronger version," "the causal version" instead of "the weaker version." But the logic is the same. The paragraph then does the transition to Sections 2-3 by identifying the causal questions: understanding, abduction, experiential sensitivity. This is exactly right. The paragraph could be updated to reference the practitioner-causal reading as well (the journey question), but the core logic doesn't change. Current para 7 (second paragraph Nick likes) starts: "An LLM is trained on a corpus that the discipline's evaluative practices have shaped." This is the sedimentation seed. It sets up Sections 2-4. It could stay essentially unchanged. So the rewrite really is paragraphs 1-5 → new paragraphs leading to the existing paragraphs 6-7. That's clean. One more thing. I want to make sure the new version doesn't lose anything important from the current version. Let me check: - Art analogy: retained and developed. - Individuals point (Kant, Lewis, Newton): could be cut or compressed. It's not essential to the argument. But it adds colour and makes the science/art comparison concrete. Maybe a sentence or two. - Davies: retained and developed. - Danto: retained and developed (indiscernibles become central). - Blind review: retained. - Dellsén: retained. - Sokal: retained (footnote). - Practitioners (Wittgenstein, Nietzsche, Merleau-Ponty): retained and developed (dual reading). - AI image literature reference [^1]: placeholder footnote, needs filling regardless. - The washing machine example: retained (it's a vivid illustration of constitutive thinking). I don't think anything important is lost. The new version adds: the constitutive/causal distinction, the indiscernibility framework, the dual reading of practitioner traditions, and a clearer dialectical structure. One final thought: the section title. "The Challenge from Authorship" is good. It captures what the section is about without committing to a specific version of the challenge. I'd keep it. ## Diagnosis The problem with paragraphs 1-5 is architectural, not sentential. The raw materials are almost all present — Davies, Danto, blind review, Dellsén, practitioners — but they're not connected by any argumentative spine. The reader encounters a series of related points (art analogy, individual-focus in philosophy, art ontology, blind review, practitioner traditions) without understanding how they fit together. There is no organizing distinction, no clear dialectical trajectory. The section discovers its own argument retroactively in paragraphs 5-6, where the stronger/weaker distinction finally appears. By then the reader has been wandering. The transcript from 14 April reveals what the spine should be: the constitutive/causal distinction, operationalised through the concept of indiscernibles. The constitutive challenge says even indiscernible texts would differ in philosophical standing if they had different histories. The causal challenge says the machine will never produce indiscernible texts. This is the framework that gives Section 1 its structure and ties it to the rest of the paper. Currently, the section uses "stronger version" and "weaker version" instead of "constitutive" and "causal," doesn't mention indiscernibles at all, and introduces the distinction late (paragraph 5), so it reads as an afterthought rather than the organising principle. Three further specific problems: 1. Davies and Danto are compressed into one paragraph ("[Pushing the analogy further]") with a bracket signalling that it's unfinished. They need room. Davies's identification of the artwork with the performance — the "double-faced entity" — is the conceptual source of the constitutive position. Danto's indiscernibles (the red canvas from *The Transfiguration of the Commonplace*) are the conceptual tool that separates constitutive from causal. Both need to be stated carefully enough that the reader understands the framework. 2. The practitioner traditions (Wittgenstein, Nietzsche, Merleau-Ponty) are confined to a generic paragraph plus a footnote. But the transcript shows they can be read in two ways — constitutive (the text is an appendix to a philosophical life) and causal (only someone who has lived the right life can produce the right text) — and this dual reading is one of the section's most interesting moves. It shows that the practitioner tradition doesn't just close the door; it generates specific, answerable questions of the kind Sections 2-3 address. 3. The opening ("While this sort of view has not to our knowledge been articulated explicitly") is wrong. Davies articulates a version of it for art. The practitioner traditions articulate versions for philosophy. Floridi and Zahavy articulate causal versions. The sentence undermines the section's own claim that this challenge is worth addressing. ## What Section 1 needs to do (functional requirements) 1. Establish the philosophy-in-the-text assumption (the rest of the paper depends on it) 2. Dispatch the constitutive objection (so it doesn't block the paper's question) 3. Show that the causal objection survives (this is what Sections 2-3 address) 4. Give the practitioner traditions their dual reading (so they aren't simply dismissed) 5. Make the art analogy do real conceptual work (source of the framework, not decoration) 6. Plant the sedimentation seed (for Section 4) ## Structural options ### Option 1: Intuition → Analysis → Distinction → Art apparatus → Practitioners → Blind review → Survival → Transition Open with the natural reaction a reader would have to the paper's question: surely an LLM can't "really" do philosophy, because there is no philosopher behind the text. Then analyse: what could "really" mean here? Two things. Either the philosopher is constitutive of the philosophical work (the text alone is insufficient), or the philosopher is causally necessary for producing text with the right properties. The first rejects indiscernibles; the second denies their producibility. Then bring in the art analogy as apparatus for making this distinction precise: Davies's action monism, Danto's indiscernibles. Then the practitioner traditions, read both ways. Then blind review as evidence that the constitutive version fails for philosophy. Then the survival of the causal version. Then transition. The strength of this option is that it meets the reader at their starting intuition and analyses it. The distinction emerges naturally from asking "what do you mean by 'really'?" The art material enters as conceptual apparatus, not as a free-standing discussion. The weakness is that the distinction arrives quite early (paragraph 2), before the reader has seen the evidence for either version, which might make it feel premature. ### Option 2: Art analogy → Constitutive position stated → Indiscernibility framework → Practitioner traditions → Blind review refutation → Causal survival → Transition Open with the art parallel: in philosophy of art, it is sometimes held that the artist is constitutive of the work (Davies). Something indiscernible from a Rembrandt, if it emerged from a washing machine, would not be a work of art. Danto's red canvases make the point: sameness in appearance does not settle sameness in standing. A parallel thought about philosophy would say that the text is not enough. Then: but this challenge comes in two forms, distinguishable through indiscernibles. The constitutive form rejects indiscernibles outright. The causal form denies their producibility. The practitioner traditions can be read either way. Blind review shows the constitutive form fails. The causal form survives. Transition. The strength here is that the art analogy opens with something concrete and vivid (the washing machine Rembrandt), and the conceptual tools emerge organically from the discussion of art. The weakness is that it might feel like the section is about art ontology rather than philosophy and LLMs — though this can be managed by making the parallel to philosophy explicit early. ### Option 3: Lead with blind review, then complicate Open: In contemporary analytic philosophy, work is assessed through blind review. This suggests that what matters is the text, not who made it. But this assumption can be challenged. Then: the constitutive challenge (Davies, Danto, practitioners) says the maker is part of the work. Even indiscernible texts differ in standing. The causal challenge says the maker is necessary for producing the right text. Indiscernibles are not producible. Then: blind review is evidence that the constitutive challenge fails. But the causal challenge remains. The strength is confidence — the section leads with its own best evidence and then shows why it withstands objection. The weakness is that it might feel like the section is answering a question before asking it, which makes the dialectic backward. ### Option 4: Science/Art/Philosophy spectrum Open: In science, nobody would say a theory is invalid because we don't know who the scientist is. In art, it is widely held that the maker is constitutive of the work. Where does philosophy sit? Then develop both pulls. Practitioner traditions pull toward art. Blind review pulls toward science. The constitutive/causal distinction is the tool for resolving the tension: the constitutive pull fails (blind review), but the causal pull survives. The strength is that the spectrum is immediately graspable. The weakness is that it might suggest philosophy is "between" science and art in a way that oversimplifies — and the point about individuals (Kant's ethics, Lewis's metaphysics) might draw too much attention. ## My recommendation I'd go with Option 2 (art analogy first), with elements of Option 1 (the intuition framing). Here's why: - The art analogy is not decoration. It's the source of the conceptual tools (constitutive identity, indiscernibles) that the section needs. If you lead with the art material, the philosophical framework emerges from it naturally rather than being imposed from outside. - Davies and Danto are real philosophers making real arguments. Using them gives the section authority and anchors the discussion in established literature. The current version's "imperfect comparison" framing undersells this. - The practitioner traditions are more interesting than the current version allows. The dual reading is a genuine philosophical insight: it shows that Wittgenstein-as-therapy, Nietzsche-as-drives, Merleau-Ponty-as-flesh are not simply closed doors but generate questions that the paper can answer. This deserves body text, not a footnote. - Opening with the reader's intuition (Option 1 element) can work as the section's first sentence or two, before the art analogy develops the idea. ## Specific paragraph sketch (Option 2, refined) Paragraph 1 — THE CHALLENGE, VIA ART: A natural reaction to the claim that LLMs can produce philosophy is that they cannot, because there is no philosopher behind the text. In philosophy of art, a structurally parallel thought is well established. Davies identifies the artwork with a performance that specifies a focus of appreciation — a "double-faced entity" whose visible face is the displayed object but whose equally constitutive other face is the artist's creative action. Something emerging from a washing machine that was indiscernible from a Rembrandt would not be a work of art, because the right history of making is absent. Danto's red canvases make the point from the other direction: perceptual identity does not settle identity of standing. A parallel thought about philosophy would say that the text is only half the work — the other half is the philosopher's intellectual activity. Paragraph 2 — TWO FORMS (THE DISTINCTION): But the challenge from authorship can be pressed in two structurally different ways. On what we will call the constitutive form, the philosopher stands to the philosophical text as the artist stands to the artwork in Davies's ontology: the relation is internal, so that even a text indiscernible from genuine philosophy would fail to count as philosophy if the right process were absent. On the causal form, the text is the whole work — philosophy-in-the-text — but only certain kinds of producers can generate texts with the relevant properties. The constitutive form rejects indiscernibles. The causal form denies their producibility. Paragraph 3 — PRACTITIONER TRADITIONS, DUAL READING: Several philosophical traditions support some version of this challenge. Wittgenstein treats philosophy as therapy. Nietzsche treats the philosopher as a creator of values whose work expresses drives bound to embodiment. Merleau-Ponty holds that philosophical reflection "slackens the intentional threads which attach us to the world." Each can be read constitutively or causally. On the constitutive reading, the text is an appendix to a process of self-transformation or embodied engagement: the philosophical work resides in the philosopher rather than in what is written. On the causal reading, philosophy is indeed in the text, but only someone who has undergone the relevant therapy, lived with the relevant drives, or attended through the relevant intentional threads could produce a text with the right properties. The two readings generate different challenges — and the causal reading, unlike the constitutive one, is not dismissed simply by adopting a text-focused conception of philosophy. Paragraph 4 — BLIND REVIEW (REFUTATION OF CONSTITUTIVE): In contemporary analytic philosophy, philosophical work is assessed through blind review. [Dellsén et al. on progress as ideas becoming publicly available.] [Sokal footnote.] If the relation between philosopher and text were constitutive — if what made something philosophy were partly a matter of who produced it — blind review would systematically set aside the thing that matters. The discipline does not treat it that way. For present purposes, then, work and text are one and the same. [Then the existing paragraphs 5-6 (currently 6-7), updated with "constitutive" and "causal" terminology instead of "stronger" and "weaker."] ## Things to decide 1. The individuals point (Kant's ethics, Lewis's metaphysics vs Newton, Einstein). I'd fold it into paragraph 1 as a sentence or two, or cut it. It adds colour but doesn't do enough argumentative work to be its own paragraph. 2. The Sokal footnote. Keep it. It's a counterpoint to blind review — a case where evaluation was based on the author's credentials rather than the text's quality, and the result was a published hoax. 3. How much Danto to include. The red canvas example should be mentioned by name — it's the classic case. A sentence is enough. 4. Whether to mention Enrico's paper. You could cite it for the survey of monist ontologies, but since he's a co-author of the present paper, a self-citation might be awkward. A note for you both to decide. 5. Whether the science/art comparison deserves its own paragraph or just a sentence. I lean toward a sentence — Nick's example of the philosopher who studied most-cited philosophers is excellent but might not be citable, and the point can be made briefly. 6. Whether to use the word "sedimented" in the new paragraphs or save it for the final paragraph. I'd save it. It has more force as a summary term than as a passing remark. ## What this structure achieves The section now has a clear dialectical spine: - Art analogy → constitutive position stated → indiscernibility framework → practitioner traditions with dual reading → blind review refutes constitutive → causal survives → this is what the rest of the paper addresses → the corpus preserves the relevant resources. Every paragraph leads to the next. The art analogy introduces the framework. The framework generates the distinction. The distinction structures the practitioner discussion. The practitioner discussion shows what's at stake. Blind review settles the constitutive question. What remains is the causal question — and that's the paper. ## Remaining uncertainties - I'm not sure whether paragraph 2 (the distinction itself) should be a standalone paragraph or folded into paragraph 1. If it's standalone, it makes the distinction very prominent — almost as if it's the section's main contribution. If it's folded in, it reads more naturally but might be too compressed. I lean toward standalone, because the distinction really is the conceptual move that organises everything. - I'm not sure whether the practitioner traditions should come before or after blind review. Before: they add force to the challenge, making the blind review refutation more impressive. After: they complicate a refutation the reader thought was settled, generating a richer dialectic. I went with before, because I think the section reads better if the challenges are all stated before the response begins. But the transcript suggests Enrico might prefer the complication move (stating the distinction, then showing how practitioners can be read both ways). Either works. - I'm not certain how much of the current paragraph 6 needs rewriting. The core logic is right, but the terminology ("stronger version," "weaker version") needs updating. The list of what the philosopher might contribute (understanding, abduction, experiential sensitivity) already maps onto Sections 2-3. It might need a mention of the practitioner-causal reading as a fourth item in the list, or the practitioner-causal reading might be subsumable under the others (the journey that produces understanding; the embodied engagement that produces experiential sensitivity). This is a substantive question about the paper's architecture, not just a wording issue. --- A few additional notes on specific textual issues: The footnote `[^1]` is still a placeholder: "reference the ai image literature here." If you're rewriting the section, this is where you'd cite work on AI art and creativity — the literature that denies AI images are artworks because no artist is involved. Colton and Wiggins (2012) is in the Learning folder and might be relevant, though there's probably more recent work on this. The `[Pushing the analogy further]` bracket in the current paragraph 3 confirms this is draft scaffolding, not finished prose. The current paragraph 5's footnote `[^pr]` (Overgaard, Gilbert, and Burwood) is fine as a reference note, but the philosophical substance — the dual reading — should be in the body, not buried there. *Una sezione che scopre la propria struttura solo nel penultimo paragrafo sta confessando di non averla mai avuta.*