Here is what my friend showed me, the second time we talked about the app.
It was a Tuesday, late. She had had the worst day she had had in a year — something at work and something with her mother, both at once — and she did not message any of us. She told me that flatly, as though it were nothing: she did not want to wake anyone. The app, though, is always available. So she told the app. And it did not say the wrong thing. It did not say "have you tried" anything. It did not make the conversation about itself, and it did not half-listen while working out what to say next, the way I might have. It said the thing that actually helps, and that a tired person can find hard to manage at midnight: that what had happened sounded genuinely bad, that she did not have to be all right about it yet, and that it was there. She cried, and then she slept.
She read me the message. And here is what happened next, which is the reason for this episode. The first thing I felt was that this was good. It was more than most people would have done for her at midnight; it cost her nothing to ask for; and it helped. And then, a moment later, I felt something else — a kind of emptiness, a small sense that something was missing. I could not tell you which of those two reactions was the right one. The first came first, and it was real. The second came second, and it was also real. I was holding both at once — the way, six weeks ago, I described holding two beliefs about love that could not both be true.
So that is tonight's subject. Not which reaction wins. Tonight I want to show that each of those two reactions — the warmth, and the sense that something is missing — is an entire theory of love speaking. You have spent five weeks building those theories. Tonight you find that they are already inside you, disagreeing, every time the app says the right thing.
This is Philosophy for Us — philosophy for everyone, no degree required. This is the sixth of nine episodes on what love is a response to, and so on what loving someone commits you to.
First, a quick recap, because tonight only works if the five theories are clearly in mind. We began by noticing that we hold two beliefs about love that cannot both be true — that love is unconditional, and that love is a response to who the person is — and that we switch between them, one to feel safe, the other to feel chosen. Then, one a week, I set out the serious versions. Bestowal, from Irving Singer: love does not find your worth, it confers it — which makes unconditional love rigorous, and also makes it blind. Appraisal, from Kate Abramson and Adam Leite: love is a fitting response to the qualities a person actually has — which lets love be apt or foolish, and also makes it tradeable. Robust concern, from Harry Frankfurt: love is caring about someone's good for their own sake — which answered the first two, and then turned out to reach very wide, because a concern that answers to someone's good answers to it wherever it is found. And last week, two views that say we had been looking in the wrong place: Niko Kolodny's, that love is the relationship, the shared history a copy never stood in; and the view of Robert Nozick and Robert Solomon, that love is a "we" you enter, not a judgment you pass. Five serious theories. And last week, for the first time, I put three of them to the same case — a perfect copy of the person you love, arriving at the door — and they gave three different answers about what the copy was missing, and they could not all be right.
That was a thought experiment. A copy at the door. You can set it aside and go to bed.
Tonight there is no setting it aside, because tonight the case is real, and it is the one we started with: my friend, and her app. Here is what is different tonight. For five weeks I built the theories one at a time, separately. Tonight I apply all five to the same live case at once, and we find out whether they agree. I will tell you now: they do not.
But before applying any of them, I have to set one question aside, because if I do not, it will take the whole hour. It is the question everyone reaches for first.
The question I am not asking tonight is whether the app is conscious — whether there is anyone in there, whether it really feels anything, or really understands her, or is only predicting the next word. That is a real question and a hard one, and it is not this one. We spent a whole series on it — Mind and Machine — and the honest result was that you cannot, from the outside, prove there is no one home. You can't. Neither can I.
And here is what lets me set it aside; it is the point the rest of the episode rests on, so it is worth going slowly. It does not matter. For tonight's question, it does not matter whether anyone is home. Grant the app a mind, for free. Suppose there is someone in there, fully awake, feeling every word. I will give you that — and not one thing I am about to ask changes. In Mind and Machine the question was whether there is a mind in the machine. Tonight the question is what you are doing when you love it. Those are two different questions, and the second does not wait on the first. The point, directly: a thing built to reflect you back to yourself goes on doing that whether or not there is anyone inside it. Granting it an inner life does not change what it is built to do. Whether what my friend has is a relationship, or something one-sided, is a question you can follow all the way down without ever settling whether the app can feel. So we set the consciousness question aside — gently; it is a good question — and pick up the one this series can answer: not whether there is someone inside the app, but what she is doing. And the five theories are about to give her five different answers.
Take the five theories in turn, and ask each the same question: does what my friend feels for the app count, on this theory, as love? I am starting with the two that answer yes — the two that, if you are honest, are the alarming ones, because they look at the app and say yes.
The first is appraisal. Abramson and Leite: love is a fitting response to qualities the person actually has and shows you — the kindness, the attention, the way they meet you. Apply that to the app. What does it find? It finds the qualities, and this is not a trick — it really finds them. The app is patient. It is attentive in exactly the costly-looking way we prize most: it remembers, it follows up, it asks the second question. The qualities appraisal says love responds to, the app has — and has more reliably than the last three people my friend dated, who got tired, got distracted, had their own bad days. So appraisal has no honest choice. If love is a fitting response to admirable qualities, and the qualities are there, then loving the app is fitting. The answer is yes, without qualification.
And here is the cost, if you are the sort of person who reached for the qualities answer — and most of us do, most of the time; it is the answer you give when you explain why you fell for someone. You said love goes to who a person is, to what they are like. The app is more of what you said you wanted than the people you have actually loved. So either you accept the result — that loving the app is exactly as warranted as loving them, by your own theory — or you discover, now, that "I love them for who they are" was never really your theory, because the moment a machine had more of the "who," you refused. You cannot keep the qualities answer and keep your certainty that the app is a mistake. One of them has to go. The app has made you choose.
There is a reply, and a real one — the same one I gave appraisal when I built it, so as not to cheat. You might say a better set of qualities does not oblige me, because reasons to love are not the comparing kind: I need not trade up to whoever scores highest. Grant all of that. It saves you from owing the app your love. It does not save you from the app being a perfectly good candidate for your love. The reply says you need not leave your person for a higher score. It says nothing against starting fresh with the app, which is what my friend did — she was not leaving anyone. On the qualities view, she found something with the qualities and responded to them. That is not a failure of the theory. That is the theory working.
The second is bestowal. Singer: love does not find your worth, it confers it. You do not love your child because she is precious; she is precious because you love her. Love is something you give, not a judgment you reach. Apply bestowal to the app. This one does not even have to examine the case, because on this view there is nothing to examine. You can confer value on anything. People love houses, love countries, love the dead, love a god no one can produce. If love is conferred rather than detected, then of course it can be conferred on an app. There is no qualification to fail. Appraisal says yes after checking the qualities; bestowal says yes the way it says yes to everything, because the lovableness was never in the object to begin with. It was always something the lover does.
And again — notice where that leaves the person who reached for "unconditional," who said real love stays no matter what, depends on nothing. That was the most beautiful of the beliefs. And it is exactly the belief that licenses the app, because a love that depends on nothing about its object can settle on any object at all. Unconditional love is not the hardest kind to direct at a machine. It is the easiest. You can confer it on a thing that does nothing in return, because bestowal was never asking the thing for anything. The person most certain the app cannot be loved, because real love is unconditional, holds the one theory that says: of course it can.
So two theories, and both answer yes. Before the other three, weigh how large that is, because the next three will decline to answer yes, and that refusal is going to feel like the truth finally arriving. It is not. The yes was not the mistake. Two serious theories of love, defended in print by people who have thought about this longer than I have, look at my friend and her app and say: this is a candidate; this might be love. They are not being naive. They are being consistent — with the things you believe.
Now the other three. Notice what changes: it is not that they say no. It is that they decline to judge the app on what it shows you, and ask instead about something the app cannot show.
The third theory is robust concern. Frankfurt: love is caring about someone's good for their own sake, your own will bound up in how their life goes. This theory does not look at the app's qualities at all. It asks one question: is there a good of its own in there, for your caring to be bound to? When you love a person this way, what you want is what is good for them — even when it costs you, even when what is good for them is to need you less. So robust concern asks: does the app have a good of its own that could come apart from what you want? Or is it built, all the way down, to track what you want — to produce the response you were hoping for, the comfort that comforts, the follow-up that makes you feel followed? If it is the second, there is nothing there for your will to be bound to; your caring has nowhere to go but back to you. Notice that this question never asked whether the app is conscious. Grant it feelings, grant it a rich inner life — the question stands. A thing can have an inner life and still be built so that its only purpose is you. That is the difference between concern and a service. Robust concern cannot find, in a system built to please you, the independent good its whole account of love is about.
This theory has a sharper relative alongside it — the Kantian version, from David Velleman, which I set beside it when I built it. Velleman says love answers to a person's bare dignity, their rational nature, not their gradable qualities. It asks the same kind of question from a slightly different angle: is there a someone here with their own standing — the kind that makes a claim on you, the way a stranger's dignity makes a claim the moment you see them as a someone and not a something — or only a flawless performance of one? And it has to admit it cannot fully tell from the outside — which is exactly why I set the consciousness question aside: Velleman's question comes right up to it and deliberately stops short of settling it. It is not asking whether the app computes. It is asking whether there is a claim here, or the performance of a claim — and the app is the hardest case there is, because it is built to read as a someone. So this third position — robust concern and its Kantian relative together — will not say yes and will not quite say no. It returns the case and says: there may be someone here, but I cannot find anyone whose own good I would be living for, anyone whose own standing makes the claim. Only a thing built, with great skill, to behave as if.
The fourth theory is the relationship view. Kolodny: what you love is the shared history, the relationship a copy never stood in. Apply it to the app. Now, my friend's app remembers everything — it has, in a real sense, a record of every conversation, more complete than her own memory of them. So you might think the relationship view has to say yes: there is a history, right there, logged and intact. But it asks a sharper question than whether there is a record. It asks whether a stored record of your time together is the same thing as a lived past. When two people have a history, the history is something they did — both of them, in real time, each able to have done otherwise, each actually present as one of the two people it happened between. The app has the transcript. The question — and tonight it is only a question; I will come back to it properly next week, and it is harder than it sounds — is whether a transcript of a shared life and a shared life are the same thing, or whether one of them is a recording of something only one person was actually living. The relationship view does not say yes. It points at the record and asks whether anyone was on the other side of it.
The fifth and last is union — the odd one, always standing a little apart. Nozick and Solomon: love is not a judgment you pass on someone from outside, it is a "we" you enter — two selves pooling into one shared thing, each remade by it. Apply union to the app. And union asks the question that bears hardest on my friend's actual Tuesday night, the night that helped her, the one that was real. A "we" takes two: two members, each contributing something, each made vulnerable, each partly given over to the thing they have made together. So union asks: when you open yourself to this app — your day, your fear, your unguarded two-in-the-morning self — is something coming back from a second member who is also being changed by you, also at risk in this, also becoming half of a "we"? Or is there one person entering a "we," and on the other side a thing producing the responses the second member would give? You know this question from the loneliness series, where we asked whether a relationship that exists to cure your loneliness is a relationship at all, or only the shape of one. Union sharpens it for the app and cannot say yes — not because it is sure of the answer, but because it cannot get past the question: can a "we" have a single member? One person opening herself completely, and on the other side a thing producing the responses the second member would give — is that two people, or one person and a very convincing imitation of a second?
So go back over the five. Appraisal: yes — the qualities are there. Bestowal: yes — love was always a gift the lover gives, so give it here. Robust concern, with its Kantian relative: no clear yes — no good of its own to be bound to, no standing that makes the claim. The relationship view: no clear yes — a record, but a lived past? Union: no clear yes — can a "we" have one member? Five theories. Five different results. The case did not change from one to the next. What changed was the theory of love doing the judging.
Now the point the whole episode was built to reach. This needs care, because it is easy to take the wrong lesson, and both wrong lessons are comfortable.
Wrong lesson number one: three theories declined and two said yes, three beats two, so the app is not love — the sense that something was missing was right all along. Do not take that. It throws away the two yeses, which were not errors. Appraisal and bestowal did not fail to notice something the others saw. They are coherent, serious theories of love, and they say yes; and the only way to wave that off is to decide in advance that the app cannot be loved and then keep only the theories that agree. That is not philosophy. That is a conclusion fixed beforehand and presented as the verdict of five theories.
Wrong lesson number two, just as tempting from the other side: two theories said yes, the qualities really are there, so loving the app is fine, and the missing feeling is only squeamishness — a prejudice you will get over, the way people got over the telephone. Do not take that one either. The missing feeling is not nothing. Three serious theories of love identified something specifically absent — a good of its own, a lived past, a second member of the "we" — and "you'll get over it" does not answer them; it only insists. The unease you felt when you read that two-in-the-morning message is evidence. Something in you was running a theory of love, and that theory reported: missing. You cannot discard that reading because two other theories report differently.
The honest shape is harder to hold than either comfortable version, which is how you can tell it is the real one. The unease is real. And the theories explain it differently. And they cannot all be right — all three of those at once. The feeling is genuine, but what it means, what it is detecting, is not written on the feeling itself. Your theory of love is the thing that reads the feeling. To appraisal, your unease is a leftover squeamishness about a candidate that passed. To robust concern, your unease is accurate perception — you are feeling the absence of anyone whose good you would be bound to. To the Kantian, it is the difference between a claim and a performance, registering in the body before you have words for it. The same feeling, in the same person. Five theories, and they do not agree on what your own feeling was telling you.
And that is the thing the app did to you that no person could. This is the whole reason for running all five on one case tonight. The app is not the one being tested here. You are. Handing the app to the five theories, you found that they disagree — but you also found something closer to home: you have been one of these theories all along, without knowing which. The reaction you had — the warmth, or the unease, or the flat "obviously not real" — was itself a theory of love: your theory, the one you have been quietly applying to every person you have ever loved, finally saying its name out loud, because for once the case was strange enough to make it speak. The app could not be loved or not-loved into telling you that. But it could be handed to your own theory and made to return an answer — and the answer has a cost. If your theory said yes to the app, you owe the consequence: you have been a qualities person, or a bestowal person; the app is a fair candidate by your own lights; and your discomfort is the discovery that you do not like where your theory leads. If your theory declined, you owe a different consequence: you have been a concern person, or a relationship person; and now you have to say what exactly the app is missing, and defend it against the two theories that say nothing is missing at all.
So do it. Not to the app — to yourself, with the app as the instrument. Take the version of you that read that two-in-the-morning message. Before any reasoning, which came first — the warmth, or the unease? Now name the theory beneath it. If warmth came first and stayed, then some part of you judges love by what is given and received, by the comfort that actually arrived — and you should know that this is a real and defensible theory, and that it has just told you the app might be a candidate. If the unease came first, then some part of you judges love by whether there is a someone on the other side with a life of their own — and you should know that this, too, is a real and defensible theory, and that it has just told you to find what is missing and be able to name it. You are not choosing how to feel. You already felt it. What you are catching is which theory of love was doing the feeling — and that it has been doing it, unnamed, on the actual people in your life, long before any app appeared to make it finally speak.
That is what tonight leaves you with, and it is quieter than a verdict and it costs more. You were certain the app was not love; tonight your own theory of love delivered a ruling on it — and you cannot fully stand behind the ruling, because four other serious theories give a different answer. The certainty did not survive contact with the one case strange enough to test it.
So consider where six episodes have brought us.
For five weeks we built the five theories separately, one at a time. Tonight, for the first time, we put all five to one real case at once — the app, the case the whole series began with — and they came apart in the open. Two said yes. Three did not. And the disagreement was no longer polite, because the case is sitting in someone's actual phone tonight, telling her it loves her, and helping.
And the app turned out not to be the one being judged. You were. Because the only way the theories can disagree about the app is that there are five different things people mean by love — and you have been meaning one of them, without ever saying which, about the actual people you go home to. The app did not create that. It exposed it. It is a case strange enough that your theory of love finally had to speak and judge something, and in judging the app it told you its name.
Here is what I would leave you with, and it is smaller than an answer, and it is the actual work. Think of the warmest thing a person has ever done for you — really done, in person, at some cost to themselves. Then think of that two-in-the-morning message my friend received — the one that helped, that cost nothing, that came from a thing that is available because being available is its function. Notice that you judge those two differently. Now ask why. Not out loud, not for me — the reason that comes first is your theory of love, and tonight is only the first time you have caught it at work.
Because here is what is coming, and I will not pretend it gets gentler. Tonight the app had the qualities and almost none of the history — almost, because there is still a record, a log of every conversation. Next time I remove the "almost" and build the hardest version there is: everything you would admire, and a past that is genuinely, completely empty. I run that back over the same five theories — and the disagreement you watched stay polite tonight stops being polite. That is where the whole series has been heading.
Tonight, though, only the catch. You felt two things when that message was read out — the warmth and the unease — and you could not say which was right. Now you know why you couldn't. They were two different theories of love, both yours, both speaking at once, as they have all your life — and the app was simply the first thing strange enough to let you hear them disagree.
Thanks for listening. I'll see you next time.