Do Automatic Sweethearts Work?

A couple of weeks ago I briefly discussed, prompted by David Levy’s treatment of the issue in his book Love and Sex with Robots, whether a robot can be said to love a person if they say they do and act as if they did. Today I’d like to continue this discussion.

In his 1909 book The Meaning of Truth, the great William James asserts that a statement is only meaningful if it makes a practical difference whether or not it is true: “if it can make no practical difference whether a given statement be true or false, then the statement has no real meaning.” (p. 52) However, in a footnote later in the same book (p. 189), he corrects a claim that he made in his previous book, Pragmatism, where he declared the terms ‘God’ and ‘matter’ for synonymous “so long as no differing future consequences were deducible from the two conceptions”. Now, however, he no longer believes this, because even if the godless universe were exactly like one in which God does exist, believing the one or the other would definitely make a difference for us. “Even if matter could do every outward thing that God does, the idea of it would not work as satisfactorily, because the chief call for a God on modern man’s part is for a being who will inwardly recognise them and judge them sympathetically.” James then asks us to consider an analogy which he thinks will convince us that there is indeed a relevant, meaningful difference between the two hypotheses:

“The flaw was evident when, as a case analogous to that of a godless universe, I thought of what I called an ‘automatic sweetheart,’ meaning a soulless body which should be absolutely indistinguishable from a spiritually animated maiden, laughing, talking, blushing, nursing us, and performing all feminine offices as tactfully and sweetly as if a soul were in her. Would any one regard her as a full equivalent? Certainly not, and why? Because, framed as we are, our egoism craves above all things inward sympathy and recognition, love and admiration. The outward treatment is valued mainly as an expression, as a manifestation of the accompanying consciousness believed in. Pragmatically, then, belief in the automatic sweetheart would not work, and in point of fact no one treats is as a serious hypothesis.”

Yet just a year later, in December 1910, the philosopher Edgar Arthur Singer gave an address before the American Philosophical Association at Princeton, entitled “Mind as an Observable Object” (later published as the first chapter of his 1924 book Mind as Behavior), in which he directly attacks James for his alleged inconsistency. Pragmatically, a soulless person (that is, one that lacks subjectivity and any form of mental awareness) should be regarded as fully equivalent to the usual kind, to a person with a soul. Singer insists that it would not make any difference whatsoever whether the other really feels anything at all or just behaves in a way that is consistent with real feelings, that is, in such a way that we cannot detect any difference between what they do and what a real, conscious and self-aware person would do. Thus, contrary to what James suggests, for all intents and purposes an automatic sweetheart is just as good as a real human lover.

When we occasionally call a lover “soulless”, we do, according to Singer, in fact refer to a certain (already observed or predicted) behaviour, so if there is a difference between the soulful and the soulless it is a difference in behaviour: “If I imagine myself come to believe that my mistress, with all her loveliness, is really without soul, I cannot think what I should mean by this if it be not that I fear her future conduct will not bear out my expectations regarding her. Some trait or gesture, a mere tightening of the lips, hardening of the eye, stifling of a yawn, one of those things we say are rather felt than seen, would have raised in my mind the suspicion that she might not to my fuller experience of her remain indistinguishable from a spiritually minded maiden.” If the distinction between ‘soulless’ and ‘soulful’ means anything, then it is this. “Consciousness is not something inferred from behavior; it is behavior.”

James’s point, of course, was that we wouldn’t be happy with a lover of whom we knew that they didn’t really feel anything for us and that all their seemingly loving actions deceive us to the extent that they indicate some kind of emotional involvement on the part of our lover. Yet Singer could respond that we might well be unhappy with an automatic sweetheart, but that we really shouldn’t be because to react like that is completely irrational, given that a real human lover would do nothing different from the automatic one.

It is interesting, though, to see how neatly Singer’s description of a “soulless” lover (where the term can be meaningfully ascribed) fits with the descriptions that we find in literature of equally unsatisfying women and with the accompanying eulogies on the virtues of the artificial lover (as, for example, in Ovid’s Pygmalion, Hoffmann’s The Sandman, or Villiers’ The Future Eve). Once again, it is the real human lover who is decried as soulless, the one that turns out not to be completely reliable, completely with us, completely there for us. It is the yawn that indicates the lack of soul, a less than interested gaze. That is the danger that always exists when we risk getting involved with real human beings. They might lose interest in us, might grow cold and unresponsive, might stop loving us. If that is an indicator of soullessness, then each and every one of us is soulless, and only an automatic sweetheart, one whose eyes will always gaze lovingly at us and will never lose their shine, whose lips never tighten, but are always soft and welcoming, and who will never have to stifle a yawn, only such a one can be said to have a soul.

Thus it appears that the effect of denying that there is any difference between a real person and a fake person, between a real human lover and an automatic sweetheart, is that the soulless becomes, or comes to be regarded as, the truly soulful, and the soulful the truly soulless.

Do robots love you if they tell you so?

I just finished David Levy’s Love & Sex with Robots, which was first published in 2007 (HarperCollins), and which interested me because of my current research focus on posthumanism in general, and more specifically on contemporary visions of our posthuman future. The book is fairly predictable and its content can be summarised in one sentence: we will, very soon, love robots and have sex with them, and that is absolutely fine and in fact to be welcomed since it will solve a lot of relationship problems that we regularly suffer from today.

As is customary among those who believe in the power of man-made technology to eventually “achieve all things possible” (F. Bacon), Levy seems to have no doubt that robots will soon be able to think and feel just as we do, or most likely even better than we do.“The robots of the mid-twenty-first century will (...) possess human-like or superhuman-like consciousness and emotions.” (10) Will they really? How can we possibly know that? Yet Levy has little patience for people who doubt his confident assertions. He compares strong AI sceptics to those Christian fundamentalists who refused to accept human evolution as a fact and to those who insisted despite all evidence to the contrary that the earth was flat (21). And just as Darwin and Galilei have been vindicated, Levy believes he will too, when we will have seen all his predictions come true. Except that in Levy’s case there is not a shred of evidence that machines will soon be able to think and feel. It is not a matter of ignoring the evidence. There is no evidence. All we have managed to achieve yet, and all we are likely to achieve in the foreseeable future, is the creations of machines that can appear to be conscious and to possess certain emotions. 

Levy spends many pages of his book providing evidence that humans have a strong tendency to perceive and treat inanimate objects as living, conscious agents even when they know that they are not really conscious or alive. And people can fall in love with the strangest things, even computers. But all that proves, if it proves anything, is that we are easily duped. It may indeed turn out that once we are able to build robots that are sufficiently convincing in their appearance and behaviour we will find it very difficult not to attribute consciousness to them when we interact with them. But that doesn’t mean that they are conscious, or that we are justified in attributing consciousness to them.

However, Levy disagrees. For him, the appearance of consciousness is not only in all practical matters just as good as actual consciousness (the pragmatist approach), but it actually is one and the same thing (the logical behaviourist approach): the (behavioural) appearance of consciousness is consciousness. “There are those who doubt that we can reasonably ascribe feelings to robots, but if a robot behaves as though it has feelings, can we reasonably argue that it does not? If a robot’s artificial emotions prompt it to say things such as ‘I love you,’ surely we should be willing to accept these statements at face value, provided that the robot’s other behavior patterns back them up.” (11) But why “surely”? It does, after all, seem to make sense to distinguish between someone who merely says that they love us and someone who really does. But of course Levy’s point is that the only way we can judge whether someone really loves us is by analysing their behaviour. The fact that somebody verbally declares their love for us might not be sufficient to attribute real love to them, but if in addition they are always there for us, listen and talk to us, look after us, always cover (and scratch) our backs, kiss and embrace and caress us and have sex with us whenever we need or want it, then we would be hard-pressed to deny that they really love us. If they do everything that we can reasonably expect anybody who really loves us to do, then it is hard to see what it can possibly mean to say that, despite all, they don’t really love us. And if we cannot find a real (meaningful) difference between a human person who loves somebody and one who consistently and permanently behaves as if they did, then why should there be such a difference when the one doing the loving is not a human person, but a robot?

It seems to me, though, that there is indeed an important difference between the two cases. If a human being behaves in all respects consistently and constantly just as someone would behave if they really loved us, then by far the best explanation for their behaviour is that they really do love us. It just doesn’t seem possible that somebody who does not love us would always behave to us in a manner consistent with real love. We would expect them to show their lack of real love in some way. It need not be something obvious, and it need not be obvious to us, but we would expect there to be something that distinguishes the behaviour of the person who really loves us from the one who only pretends to do so. It would be nothing short of a miracle if a pretended lover would, throughout his life, act exactly like a real one, precisely because such behaviour would be entirely inexplicable.

This, however, is not the case with robots. If they behave in all respects exactly like we would expect someone to behave who really loved us, then we have a perfectly good explanation for why they behave like that, namely that they have been designed that way. Levy claims that our knowledge that robots have been designed to manipulate us into believing that they really love us is irrelevant and should make no difference to us: “Even though we know that a robot has been designed to express whatever feelings or statements of love we witness from it, that is surely no justification for denying that those feelings exist, no matter what the robot is made of or what we might know about how it was designed and built.” (12)

“Surely” (again)? I think it makes all the difference. Again, we might be tricked into believing that robots truly love us, but that doesn’t mean that they do. And while it might make no sense to distinguish between the real and the merely apparent when it comes to human behaviour that is consistent with the actual presence of a certain emotional disposition - simply because we would not be able to plausibly explain such behaviour -, the fact that we know the other to be a robot, that is a machine designed to behave as if they loved us, is, by providing a perfectly good explanation for such behaviour, sufficient to justify our refusal to believe that they really do.