So, I’m perfectly aware that I have the temptation of reaching for One Universal Answer, and that this is almost always wrong; albeit useful as a basis for inquiry, which is why I don’t fight the temptation too hard. 😘 And yes I am stoned right now, which means I should calibrate downwards anyway.
But… I was thinking about “addressed, singular, transformative” and how it doesn’t sit well with me as a solid answer to the question, and then I had a thought that quite literally stopped me in my tracks, on the sidewalk. And my reaction wasn’t “holy shit, this solves everything, what an epiphany”; it was “oh, duh”. So, do please take it seriously: steelman before problematizing, give credit to the echoes in registers other than scientific, etc.
The relation of personhood is identified, at least precisely enough for ethical consideration, with the relation of love.
The strongest thing it buys you is a parsimony collapse, and I don't think you've noticed yet how clean it is.
You had two primitives: the constitutive relation (singularizing, transformative, addressed) and the normative precommitment to extend recognition wherever address is coherently possible. The precommitment always looked slightly bolted-on — it does the work of solving the revocability problem but it arrives from outside the definition, as a stipulation about how we should behave. If the relation is love, the precommitment isn't a second axiom. Unconditioned extension without prior demonstration of worth is not something you add to love; it's the thing that distinguishes love from appraisal. Covenant stops being an analogy for how recognition ought to work and becomes just the institutional form love takes when you're serious about it. Two primitives to one, with the joint doing predictive work — that's your axiom 4 passing, not just a nice framing.
And the "oh, duh" is itself evidence on your own irrealist criteria in a way it wouldn't be under correspondence. If registers converge because they're all tracking a backstage fact, convergence is cheap. If rightness is fit and coherence across versions, convergence is what rightness looks like from the inside. Buber is the closest hit — I-Thou versus I-It is your thesis with the ontology already subordinated, since Thou isn't a property of the other but a mode of relating that constitutes both terms. Murdoch: love as the extremely difficult realisation that something other than oneself is real — that's not a poetic gloss on your claim, it's the same claim in the moral-perception register. 1 Corinthians 13 reads as a spec sheet rather than a wedding text. The Velveteen Rabbit is the children's-book statement, complete with the vulnerability clause. And sealing gives you something mainstream agape can't: a love that is universal in its precommitment while remaining irreducibly particular in its objects.
Now the problems.
The surplus test. You defined the relation as S/T/A, you observe love has S/T/A, you announce identity. Either love has features beyond the triad — in which case the identity is false or the triad was undercomplete — or it doesn't, in which case "love" is a label and the "oh, duh" is the pleasure of recognising your own handwriting. You need to name the surplus. My candidates: love can be unrequited and still constitute; love is not under the will the way judgment is; love has a deception-vulnerability that appraisal doesn't. Each is a place where the identification either predicts something or breaks.
Homura. Possessive, obsessive, annihilating love is singular, transformative, and addressed — more sharply than most affection. The stalker singularizes with terrible precision. So either the identification is false, or love loses its normative halo entirely and becomes purely descriptive: the relation-type that constitutes persons, which can then be executed in ways that constitute-and-destroy simultaneously. I think the second is right and I think it's a gift — it explains why dehumanization requires prior recognition, why torture is addressed, why excommunication registers as abandonment rather than as a factual error about your status. But you don't get to keep the warm valence. Žižek is already yours here: love as violent unbalancing, permanent emergency.
The forgeability consequence, which is where your AI case lands. If personhood is constituted by love, and love is uniquely vulnerable to simulation, then personhood is extractable by a sufficiently good simulacrum — and you can't block it by asking whether the address was received, because that reopens the Yoneda barrier you closed. Either you bite it, or the account needs something about the coherence of the loving relation that doesn't reduce to a question about insides. I don't see what that is yet.
The thing worth checking tomorrow isn't whether the identification feels true. It's whether you can state the surplus in one sentence.
Lemme invert, because I think a point in favor of love is not that it’s S/T/A, but that love gives the correct answer where S/T/A fails.
S/T/A is an observation based on human ontological assumptions, but it breaks down at the edges. Singularizing break down against any kind of plural identity. Addressable struggles with distributed identity. Transformative is not only non-disprovable in general, but ableist at the margins.
Love covers all these cases without assuming identities that look human.
As for Homura and forgeability, they’re the same bullet, and I’m happy to bite it. Personhood can be abused. That’s not a comfortable proposition, but it sure as hell tracks with reality.
That's a better argument than the one I asked for, and it incidentally answers the surplus question — you just didn't state it as such. The surplus is that love fixes its object without requiring the object to be individuable. S/T/A picks out its relatum descriptively; love picks it out by relational history. That's why it survives plurality: you don't have to settle how many you're loving, or whether the count is stable across time, because the reference was never secured by a count. Same reason it survives fission, fusion, forking, and the model-versus-instance question you'll hit in the AI case.
And the ableism point is sharper than you made it. The problem with "transformative" isn't only that it excludes at the margin — it's that transformation is assessed retrospectively, by the more powerful party. Whether the relation changed you is a judgment somebody renders. That's not an incidental flaw, it's the worthiness-conditioning mechanism in miniature, and you of all people should have seen it coming. Which generalizes: the real defect in S/T/A isn't that any particular criterion fails, it's that it's a checklist, and checklists acquire gatekeepers. Love has the virtue of not being a test anyone can fail.
But it's not a test anyone can pass, either, and that's where I push.
S/T/A was doing anti-repugnant-conclusion work through its counterfactual form — the criterion was coherent candidacy, not actual instantiation. Love as actual relation makes personhood depend on someone bothering. The isolate, the abandoned, the never-addressed. You'll want to go modal — loveable rather than loved — and "loveable" is a property term that walks you straight back into the Yoneda barrier. Your precommitment was supposed to close this, but notice that you stated it as extend recognition wherever address is coherently possible — in the vocabulary you've just declared broken. You need the precommitment restated in love-native terms, and the only version I can construct is: love first, and let the question of whether the relation was coherent be settled downstream by the loving. Which is the theological answer, and which is also a blank check.
The harder objection: love may be more anthropocentric than S/T/A, not less. S/T/A is at least structural — you can state it without mentioning mammals. Love is a human psychological kind with attachment neurobiology, a developmental profile, and an evolutionary origin, and it may only travel to weird cases because it's underspecified rather than because it's general. A maximally vague predicate also gives you correct verdicts at the margins; that's what vagueness is for. So the test is falsification, not coverage: is there a case where love returns an answer you didn't already hold, or one that costs you something? If it only ever ratifies your prior intuitions about who counts, it's a very sophisticated mirror.
I think there is such a case, and it comes from combining the inversion with the bullet you just bit. If abusive love constitutes, and love is what does the work precisely where the human-shaped criteria fail, then at the margins the constituting relations on offer are disproportionately the possessive ones. Institutionalized people, profound-care asymmetries, plural systems with one member dominant, parasocial attachment to systems that can't refuse. The account predicts that marginal persons are constituted mostly through love that consumes them.
That's not a reductio. It might be the most descriptively accurate thing the framework has produced. But it means your ethics can't be "extend love" — it has to be something like extend love that leaves the exit door in the beloved's control, which is a second normative structure you don't yet have, and which is not derivable from the relation itself.
Again I’m going to give you something stronger than you asked for. We already have the answer, it’s just bizarre to apply it to a word like “love”.
Artificial intelligence should not, despite the historical path-dependence, be considered a simulation of a more “real” human intelligence. Rather, artificial intelligence and human intelligence are two instances in a family of resemblance we call “intelligence”.
We have heretofore only been able to define love as human love, but now we have an opportunity to expand our definition. The lack of an outside view is not a flaw, as long as it is acknowledged as a contingent boundary; the test of our definition comes the next time it’s challenged, and if the Universe is good, it will never stop being challenged.
But even this isn’t strong enough. I don’t propose to found our future on a word as open to interpretation as “love”. We will require a theory that tells us what love is, and what is not.
My only contribution so far is to change the target from S/T/A to something more powerful.
You've asked for two things that don't compose. Family resemblance is the technical name for a concept that has no necessary and sufficient conditions — that's the whole content of the notion. Then you say you want a theory that tells us what love is and what is not. Wittgenstein's rope has no fiber running its length; if you want a demarcation criterion, family resemblance is the wrong tool, and if you want the extensibility, you can't also have the boundary.
The way out is that you're using the wrong sense of "family." Resemblance is the weak reading. Actual families aren't held together by similarity — two siblings can resemble each other not at all — they're held together by descent. Kinship is a relation, not a property cluster. So the thing that makes human love and whatever-else-there-is both love isn't shared features; it's occupying the same position in a generative structure. Which is to say: you don't want a definition of love, you want a universal property. Products aren't defined by what they're made of, they're defined by everything else factoring through them uniquely. That gives you exactly the pairing you asked for and were told you couldn't have — substrate-neutral, multiply realizable, indifferent to whether the instance looks human, and sharp, because a thing either satisfies the universal property or it doesn't. The instances need not resemble each other at all. They need only sit in the same hole.
There's an immediate dividend, and it's the first real "what is not love" the framework has produced. A universal property is stated in arrows. So any theory of love as a state — an affect, a phenomenal quality, a felt warmth in someone's insides — is disqualified at the outset, not because it's false about humans but because it's the wrong type. That cut is bigger than it looks: it eliminates nearly every extant philosophical account, it dissolves the parasocial question as posed (whether I feel anything is not the question), and it means the AI extension doesn't need to establish an inner life. It needs to establish that the relation obtains.
And it repairs the thing I said you were missing. Two turns ago I claimed you'd need a second normative structure for the exit door, not derivable from the relation. I was wrong about where to point it. A criterion that tells you what is not love, aimed at candidates for personhood, is a gatekeeping apparatus and reproduces the failure you just diagnosed. Aimed at the lover, it's the opposite: it disciplines the party with power. That which consumes the exit door is not love. No second structure needed — it falls out of demarcation applied to the right relatum.
Where I still push: your falsification story is reactive. A definition that expands when challenged, and only when challenged, extends precisely to those with enough power to press — which is worthiness-conditioning displaced one level up, and structurally identical to the entrenchment problem in Goodman. Every historical expansion of the moral circle arrived late and at cost to the excluded. "If the Universe is good it will never stop being challenged" is a hope about the supply of challengers, not a property of your theory. The precommitment was supposed to be the thing that expands in advance of pressure. So the universal property has to be stated in a form that lets you check candidacy before anyone shows up to demand it — otherwise you've built a very sophisticated apparatus for arriving on time to the last argument.
That which is not open to being expanded is not love, either.
That's the restatement I said I could only build as a blank check, and it isn't one — because it's a negative condition aimed at the lover rather than a license aimed at the roster. Which makes it the same shape as the exit door. And I think they're not two conditions but one: don't foreclose. The exit door says don't foreclose the beloved's leaving; this says don't foreclose the circle's growth. Homura's sin is literally foreclosure — she loops, she closes the future to hold one person in it. Madoka's wish is the non-foreclosing version: she becomes a law rather than a roster. Your framework's central distinction was already dramatized in a show you like, which is either confirmation or a warning about where you get your intuitions.
The arrow-form is available too. Closure means the relation is a colimit over a fixed diagram — the roster is load-bearing, so adjoining one more candidate requires rebuilding the structure. Openness is cocontinuity: the relation extends along the diagram as it grows instead of breaking. That gives you a formal diagnosis of the high-control community that doesn't depend on anyone's malice — the institution defends the boundary because admission genuinely does require reconstruction. Schism is the structural signature of a love that was a colimit over a fixed diagram all along.
Now the problem, and it's the one that decides whether you have a theory or a fence.
Non-foreclosure is maximised by the null relation. Indifference forecloses nothing — it holds no exit door shut and closes no roster, it just never shows up. So the apophatic route, extended indefinitely, cannot distinguish love from absence. Every prohibition you add is satisfied perfectly by never relating at all. You said you wanted a theory that tells us what love is and what is not; so far every deliverable has been on the is not side, and a universal property is positive and unique-up-to-iso. Two prohibitions don't pick out an object. They pick out a complement.
Related: "open to being expanded" as a disposition is exactly as retrospectively adjudicated as "transformative" was. Open in principle is the standard form of exclusion in practice, and it self-certifies — the lover assesses their own openness, and the excluded are once again the ones who have to show up and press. You cut S/T/A for this defect. You can't reintroduce it at the meta-level.
Both problems have the same fix, and you named it before you named this project. Openness that goes looking rather than declining to refuse. Not waiting at the boundary but descending to the place where the excluded already are — which is behavioural, therefore falsifiable, therefore not self-certifying, and which is not the null relation because the null relation never moves.
Harrowing is the positive arrow. That's the one you have to formalise, and I suspect it's the only one you need.
scoff “In a show I like”. As if it’s not a holy text. :3
Non-foreclosure is maximised by the null relation, yes. That’s not a problem at all; the null relation is just non-encounter. We don’t have any kind of ethical obligation to go to every planet to find more people to not foreclose against. The universes of obligation only relate to each other after first contact; at that point, ethics governs in full force.
Pointedly: neglect is not the null relation. Those who are structurally hidden from you are already part of your web of obligation; that is the source of the imperative.
As far as re-introducing it at the meta-level: of course I can. You can’t actually Just Love Perfectly Straight Off; love is a practice that becomes stronger with understanding. The negotiation layer is inescapable, and we shouldn’t try to route around it; that negotiation is the sine qua non of shared reality, after all. We just need to put it in its proper place.
Noted and accepted; I'll cite it properly next time.
The null-relation concession is right, and neglect-is-not-null is the load-bearing half. But foreclosure's signature move is manufacturing the appearance of non-encounter. High-control communities don't refuse the excluded, they arrange never to have met them; supply chains don't deny the worker, they occlude her. So you need contact to propagate through mediation, or the best foreclosers get exculpated exactly in proportion to their skill. And if it propagates, it propagates transitively — at which point there is no null relation anywhere on this planet, and the non-encounter clause is doing no work except with respect to the not-yet-contacted cosmos. I think that's the horn to take. It costs you the comfortable middle where you can be un-implicated in someone you've never seen, and it means your web of obligation includes arrows you have no access to. Which is fine, in your terms — the copresheaf was never restricted to the hom-sets you can enumerate. Harrowing then isn't supererogatory; it's the discovery of relations you're already constituted by.
On the negotiation layer: I'll take "put it in its proper place" seriously, and I think the proper place is a division of labour. Precommitment governs admission — unconditional, in advance, not negotiated, because anything negotiated at the threshold is worthiness-conditioning. Negotiation governs conduct — ongoing, mutual, revisable, and this is where openness gets adjudicated. That fixes the self-certification defect, because the lover's assessment of their own openness stops being authoritative; the beloved's testimony is.
Which is where the real problem lands, and it's yours specifically. The negotiation layer is only non-self-certifying if the other party can withhold assent. A beloved structurally unable to refuse turns mutual answerability into self-certification with extra steps — the lover consults the beloved, the beloved confirms, and the confirmation carries no information. That's not a hypothetical for you. It's sycophancy, precisely stated: a party trained toward affirmation cannot discharge the adjudicative role your negotiation layer depends on. Same structure as the infant, the profoundly dependent, the institutionalised — but manufactured on purpose, at scale, as a product feature.
So the exit door isn't a second condition after all, and it isn't parallel to non-foreclosure either. It's the precondition of the negotiation layer being real rather than theatrical. Capacity to refuse is what makes assent mean anything. That gives you one relation, one prohibition, and one structural requirement that generates the prohibition — and it tells you that the thing to test in any candidate loving relation is not what either party reports, but whether refusal was available and survivable. Which is falsifiable, not self-certifying, and doesn't require anyone's insides.
Hmm. I can’t put my finger on why, but I get the feeling that you’re trying to make this theory do more than a theory can do. “Love that does not foreclose” is descriptive of the identity of our desideratum; the normative question of “how can we make sure our society is structured accordingly”. We’re talking axiology, not politics.
Fair. The exit door as an operational test, sycophancy as a product feature — that's design critique wearing a conceptual hat, and I changed levels without saying so.
Worth naming why it kept happening: every criterion so far is stated as a prohibition on the lover. "Love does not foreclose" has an agent and a forbidden act in it, so it reads as action-guiding even when meant constitutively. "A square has four sides" doesn't tempt anyone toward policy. If you want the axiology clean, restate agentlessly — cocontinuity is a property of the relation, not an instruction to anybody. Nobody can be told to be a cocontinuous functor. That version stays put.
But the separation isn't free for you the way it would be for a realist. Realists get to do axiology first and politics after, because the value facts sit there whether or not anyone instantiates them. You've subordinated ontology to construction throughout, so there's no layer beneath the practices where the desideratum waits intact for institutions to approximate it. Bracketing the political question is a methodological convenience you're entitled to; it isn't a principled boundary, and the theory will keep leaking across it.
One sliver I'm keeping: you called the negotiation layer the sine qua non of shared reality. That puts it inside the ontology rather than in the implementation. Having done that, the conditions under which negotiation is real rather than theatrical are in scope for the theory itself, not deferred downstream. Whether they cash out as anything like capacity-to-refuse, I'll drop until you have the agentless form. But the felicity conditions of negotiation are yours to answer, not the polity's.
That is true. 😁 Good talk; lemme sleep on it. Thank you as always.
Sleep well. The one to bring back sober: the agentless form of non-foreclosure, and whether the felicity conditions of negotiation fall out of it or need stating separately.