Under the title, the Autonomy of Mind, I am going to discuss two theses. First a negative and contentious one, and then a positive, perhaps less contentious but rather more cloudy, one. The negative one I think in this place should be reckoned as being largely addressed against David Hume, who put himself out as the Newton of the mind and was going to give us some very clear rules governing the operation of the human mind. What I am going to offer should be seen, as it were, as another essay on the intellectual powers of man.
I want to contrast certain different types of reasoning, different types of understanding. And since intellectuals seldom tell the truth except when they are trying to be rude, I want you to go for a moment, in thought, to the Senate of a university—not here of course—but another senate where the professors are quarrelling. Quarrelling about money. Each professor wants a rather bigger share of the available money to expand his own Department, because each professor knows that his own subject is best; and behind the interchange of incivilities certain themes will come out and particularly one between the arts men and the scientists. The scientists will complain that the arts men are sloppy and subjective; they waffle away; they are little better, really, than journalists, and you have got no clear sense that what they are doing can be established as clearly and objectively true. As one particular case comes up; who shall be, say, a Reader in Ecclesiastical History—x or y? They are both going to be writing about the covenanting movement in South West Scotland, x is an episcopalian and y is a presbyterian, and this matters: whereas there is no such thing as Episcopalian, Presbyterian, Wee Free or even Humanist, Chemistry. The arts men, in return, complain that the scientists are terribly narrow, or, if they are very young, they say that they are irrelevant. They keep on plodding along, examining the sex-life of woodlice, answering questions which nobody really wants to ask, and failing to see the point of this objection and that objection. Unable to take the longer view or without the larger vision, and terribly much lacking in intuitive insight.
Well, I shall not develop this very much, partly because as I came up on the plane I found that it was done very much better in a book I was reading, G. H. von Wright's Explanation and Understanding, (Routledge and Kegan Paul, 1971) and the first chapter foreclosed a lot of what I was going to say by doing it better. Partly, also, because this distinction seems now to be conceded by all of us here; and rather than say things which are agreeable, my function is to say things which are disagreeable, and I am going to try to pick a quarrel with the other three which is not whether there is a distinction but whether this distinction is irreducible. After all, it is a long time since it was originally noticed—Plato in the Phaedrus (263a) draws a distinction something Like this in terms of a decision-procedure. Controversies about whether someone is good or just, he says, go on for ever, whereas whether something is made of iron or silver, is a matter which can be settled definitively. Pascal drew a distinction between l'esprit de finesse, and l'esprit de géométrie. In the last century Newman in this country, and Droysen and Dilthey in Germany, kept on reiterating that there was another sort of understanding besides that which could be squeezed into Hume's canons or those of the purely mathematical sciences. Nevertheless, this penny has not dropped. Even when I was an undergraduate I was made to read a very important and influential work on the philosophy of history by an American philosopher, Hempel, where he was squeezing history into what he called the ‘Covering Law’ scheme of explanation, and I am going to argue that history can't be squeezed—not simply by pointing out, which is true of course, that the results after the squeezing are not recognisable as history, but by a more a priori argument.
The key word which will separate me from the other three is the word ‘algorithm’. Waddington with only a very little prodding will produce algorithms in his account of how embryos develop, and Longuet-Higgins believes that his computers can do everything because everything that can be done can be done according to an algorithm, and yesterday Kenny—I quote—said that ‘if it is an inference at all it must work in every case’. Now I may be misquoting, or misquoting his intention, but there is here a certain ideal of being able to have a decision procedure, a method, a procedure of deciding whether something is true or false which will apply in every case. This is true of the sciences, but it is very typically not true of the humanities. We don't think that we can decide whether, if Hannibal had marched on Rome, Rome would have fallen, by looking for some universal law which applies to all Rome-like cities being marched upon by Hannibaline armies. The importance for our present concerns is whether we can have every decision being able to be decided by some decision procedure, whether every problem can be decided by an algorithm.
Some of you have already indicated to me that they rather hoped that I was going to prove Gödel's theorem. I am not going to do that, partly because it is very difficult, but more because it is exceedingly dull and takes about forty-seven pages of close, usually Germanic, print. But I shall explain what to do if you meet a computer one dark night on Forrest Hill. The first thing is to get into conversation. We have our doubts whether computers can talk, but Longuet-Higgins assures us that they can, and it is always a good thing to ‘jaw jaw rather than war war’. Engage him in a Socratic dialogue and ask him various questions. First of all, various questions to make sure that he is really a computer and not Longuet-Higgins in disguise. The crucial one is, does he operate algorithmically—that is to say, does he have just simply a certain set of very definite instructions, according to which he will answer any question that you put? If he answers ‘yes’ to this, then he is a computer, but also then you can start numbering his instructions and any combination of these. There can't be more than ℵ0 (the smallest sort of infinity that there are) instructions according to this principle; and therefore you can number them 1, 2, 3, 4 and so on up to infinity. Having got to this point, you also start to number the problems, the questions you can put to him, and the key to the whole way of dealing with computers is that they both can be numbered and therefore you can start playing off one against the other. And then you start putting more difficult questions. You ask the computer, ‘Dear computer, what is your procedure for telling that problem number n cannot be solved by procedure number n?’ And if he answers ‘7,777,777’—supposing he gives this as his answer for his procedure for telling that problem number n cannot be solved by procedure number n, then naturally the next thing to do is to ask him ‘Can problem number 7,777,777 be solved by procedure number 7,777,777?’ and if he is a very arrogant computer he may say ‘yes’. If he says ‘yes’, then you point out that that procedure is one which tells him that that problem cannot be solved by that procedure. So perhaps he is a more modest computer and will say ‘no’, and then either you could start niggling and induce him again into an inconsistency or you might be generous and allow his humility at the price of getting him to admit his humanity, and show that here is a problem that he can solve, but not by that procedure.
Now this is a tricky argument; it is a finicky argument; but it is in fact a fair one. It is one of a great range of theorems which have been discovered, ranging from Church's theorem through Gödel's theorem to Tarski's theorem. Church's theorem shows that although a very well-paid and patient computer could go on if it had infinite time and infinite money, proving all the theorems of the predicate calculus, it could not show all the well-formed formulae which were not theorems to be non-theorems. That is one end. In the middle there is Gödel's theorem, which I think is the most interesting one because it has a certain interplay between truth and proveability; and at the far end Tarski's theorem which establishes that within any formal system in which the elementary operations of arithmetic can be formalised, it is impossible to have a predicate which has the typical properties of the word ‘true’. That is, truth cannot be formalised in computer language, with the consequence that the computers in their languages, Algol, Fortran, or anything else, have to dispense with the concept of truth—And this seems to me to lead us to a certain creative theory of truth, something different from the traditional theories which are normally given—the correspondence theory and the coherence theory and the ditto theory of Strawson in his earlier years. I am not very clear exactly how this is to be worked out. What I think we do see is that if we are to give an adequate characterisation of the intellectual powers of man, then when we are concerned with what things are true, we cannot hope to be able to reduce them entirely to some set of decision-procedures—the set of algorithms which are what computers can do, and what the whole programme of formalism would require that all mental activities should be able to be reduced to.
I want now, having made this negative point, to link it up with a more positive point. That is, the negative point I've made is that some of the characteristic activities of the human mind are autonomous in the sense of not being reducible to, or representable by, purely formal logical or mathematical operations. The positive point that I want to make is certain intimations we have, most typically in moral philosophy, but extending over the whole range of our intellectual activities, about each person being in some way his own originator, his own creator of values. And here I'm going to be running against, I think, a view which is very commonly held, one which certainly goes back to Kant, a view which makes a great separation between moral philosophy and all other intellectual activities. The pure reason is thought of as being something purely academic and to be distinguished from the practical reason; and it is only the practical reason that people need really worry about. Now I want to say not this, but that all our thinking is of a piece and that what holds for our thinking about moral problems holds for our thinking about intellectual problems and vice versa. There are of course differences between different disciplines, but there is no big gulf between the one and the other.
I think the reason why people have been much more concerned with autonomy in moral philosophy is that we are all, all the time, taking moral decisions, whereas we are only some of the time and rather occasionally trying to be creative in the mathematical or the philosophical or the other intellectual sciences. Nevertheless, one can see the force of criticism even in the intellectual sciences. After all, though we are sometimes told that computers can compose symphonies, the idea of algorithmic art does seem to be one deeply counter-intuitive, and it is a criticism we often make of people, and we don't need to go to computers; we can often find writers and artists who act entirely according to the rule book, and the results they produce are very wooden. Nevertheless, it is mostly in moral philosophy that we are aware of the importance of autonomy. At the present, we will hear young men talking about the need to be authentic. They complain that their elders, (if they know French) are suffering from ‘mauvaise foi’. Or we can see exactly the same criticism being made of the Judaism of Our Lord's time. Here there was the nearest thing you could have to a moral algorithm. The Scribes and Pharisees had worked out to a very very great degree of complexity exactly how far one could go on the seventh day, exactly what one should do if one's family duties had been pre-empted by some religious obligation. Every possible question had been asked and had been answered, and there was a definitive ruling. Yet, this only produced whited sepulchres who were inwardly ravening wolves. The letter killeth, the spirit maketh alive. It is a lesson that is constantly being re-learnt and constantly being forgotten. It was the same point as was being made at the Reformation, Luther suddenly realises that monkery, going through all the hoops, is just not relevant. This is what he was saying when he was saying that justification must be by faith alone—sola fide. But within three or four generations protestantism had forgotten this lesson and had descended into what I might term the Deuteronomy of the will.
It is very easy to sense the importance of autonomy in giving an adequate characterisation of the nature of the mind. It is very difficult to give this characterisation at all clearly, and I first want to put on one side two points which are often read out of the doctrine of autonomy and which do not in fact follow. It is often taken, and it seems to be supported by a superficial reading of Kant, that autonomy is opposed to heteronomy, and that if we are to be authentic operators on our own—each man doing his own thing—then it is absolutely necessary that we should never do anything at anyone else's bidding. And I at this point went to put in a plea in praise of heteronomy. I think the best way of putting this across is to point to the virtue of loyalty. If I am loyal to someone, I show my loyalty not by always deciding myself what is the right thing to do, what's the right thing for him to do, what's the right thing for me to do, vis à vis him—but rather to be willing to accept his decisions and then going along with him. He may want to do something which I don't terribly want to do. He may even want to do something which I am not quite sure is the best thing to do—nevertheless if I am loyal I go along with him. I accept his right to lay me under obligations. Exactly the same issue turns up in, often not loyalty to a person but loyalty to an institution, where I want, rather going against the current trend, to argue that often it is the mark of loyalty and of responsible citizenship or responsible membership of an institution to be willing to act against one's better judgment. Well, this is one point that I want to clear on one side because it is one which is very often forgotten.
I want to turn now to a second point, which again very often arises, which is to think that since we ought to each make up his own mind what he is going to do, therefore anything goes. The anabaptists at Munster read out this lesson ~ obviously St Paul had been having trouble with their predecessors in Corinth; and it needs very little experience of the antinomian argument to shoot a person back firmly and squarely in the absolute conviction that not one jot nor one tittle of the moral or the legal law is to be abrogated; or again, I have noticed it often with colleagues who have had the misfortune to have come in contact with revolting students. It is marvellous what a change of mind this induces. And the important thing to see is that they are right in what they affirm; only, they are often wrong in what they forget. That is, it is not the case that the doctrine of autonomy either in morals or in matters of the intellect means that one's deciding it is so makes it so. My believing that something is true does not make it true, and my believing that something is right does not make it right. The issue of whether something is right or wrong is to be decided perhaps by me, not always but often, but it is not one that my decision thereby makes it be what my decision is.
This shows up very clearly, I think, in the intellectual case. I have been arguing that the algorithms don't answer all the questions. We can't reduce all the operations of the human mind to working according to some set of definite decision-procedures; but we should not conclude from that that in those cases where there actually is a decision-procedure, one is, nevertheless, still at liberty to decide something else. My argument against Longuet-Higgins doesn't mean that either the computer or I am at liberty to say that two and two equal five. Rather what is wrong with the doctrine of complete reducibility to decision-procedures is the completeness. We can formalise different parts of logic, and we can formalise some different sorts of legal procedure and insofar as we do it and do it well we are able to set up procedures which will enable us to see what ought to be done or what ought to be believed, but it will never be a complete job. The lesson that we should draw from the incompleteness theorems is not that formalism is always wrong but that, however far we go in laying down formal procedures for deciding different questions, there will always be other questions which will not be decided by this method. We may then be able to produce another method which will do that other problem but then there will be other problems still, which will elude both the first and the revised decision-procedure. What is wrong with the algorithmic approach is partly that it tries to prevent us asking certain questions, partly that it encourages us to ask certain other questions.
To go back to the moral case. What was wrong with the Judaism or any very well-worked-out legalism is that the question ‘what shall I do?’ is gradually eroded in favour of the question ‘what can I get away with?’ This is a very proper question for a man to ask his solicitor, not a proper question for a man to ask his confessor or pastor or any counsellor. Or to take the more intellectual case, where I am very largely inclined to blame Descartes, the question ‘is it true?’ has been replaced by the question ‘can I be sure that it is not wrong?’. Descartes, you remember, shut himself up in an airing cupboard and decided to reject all the beliefs of his elders. There is no doctrine so silly, he said, but that there has been a philosopher who has proclaimed it; and so he resolved not to accept anything other than those doctrines which he could be absolutely certain were not wrong. And this is what a decision-procedure is enabling us to do. We can do it either in the privacy of our airing cupboards or perhaps better in company with other people. Socrates arguing with Thrasymachus, or a man now trying to argue with a computer, trying to see what he can force the computer to accept. This idea of forcing, in Latin cogo, cogent—the idea of a cogent argument is a very important question. But it is not the only question. And I think it has been a great corruption of the academic world that too many academics have come to think that this was the important question. It doesn't matter, says one don, that we haven't said very much: at least it can't be wrong. And I want to say that this runs against a certain intuition we have of autonomy, that of each mind being an object on its own, guided by some idea of truth, an idea that is not subjective, not arbitrary but also not entirely external. We can't, and here I part company with Plato, entirely externalise the truth, and think of it as something set out independently of us, timeless, spaceless, and altogether impersonal. If we say this, then we are running against the intimation that we have that truth is something which has to be discovered by us, but is also something which is not made by us. It is something to which we aspire. Well, these doctrines are ones which I still find very difficult to articulate, but they seem to me to be something which is bound up with the concept of autonomy and unless we are prepared to think about ‘truth’ and about ‘argument’ and about ‘inference’ on these lines we cannot really be doing justice to a view of the world which takes seriously the existence of the mind.
Well, I don't know quite where to begin, but I think I would like to start with that dark alley in Forrest Hill where John Lucas has described himself going up and meeting an unknown creature with a great black cloth over it, and trying to decide whether it is a computer or a human being. May I invite you to consider a corresponding situation where I walk up the alley in Forrest Hill, and meet somebody who is in fact John Lucas behind this black cloth, and I try to discover whether he is a computer or a human being. And to do that I address to him the following question, which I shall write on the blackboard. Let me make sure I spell it out right.
‘Would a rational being fail to give an affirmative answer to the question on the blackboard?’
You must imagine that I am holding up a blackboard with this writing on it. Now let us just consider the situation in which this anonymous gentleman finds himself. He might give an affirmative answer: he might say ‘yes’—But if he does so, he shows that he is not a rational being, because saying ‘yes’ to this question implies that one thinks that it is rational not to give an affirmative answer to this question. So he has the option either of holding his peace or of changing the subject, in either case of course he fails to give an affirmative answer to the question at issue. And so we conclude that if he is a rational being—that is what rational beings do, they fail to give an affirmative answer—but he cannot say so: so, if he claims to be a rational being he in fact cannot with consistency answer that question. He is in no better position in fact than those poor machines upon which he pours such scorn and I want to ask John Lucas that question.
What was the question?
He has failed to give an affirmative answer.
This was a difficulty I long ago came across in trying to catch philosophical positions by their own necks, and seeing what it was that was wrong about them; it seemed that unless you had already said what it was, you did not know what it was you were questioning. And the answer to Longuet-Higgins’ attack is to point out that he has not asked a question, because what he is referring to is something which is not yet complete, until he has said what the question is. I must know if I am expected to answer, which question… and I invited him at each stage to say ‘I am very sorry, oh Computer, I do not understand these things very well. Could you say which question?’ and then he might put it, namely… Back he goes to here. Then having got to there, we have to start off again, round again. Now this is an admirable procedure if one meets a fierce Christopher Longuet-Higgins on a dark night, but it does not exert any force against me. What I've got to show, though, is not that I can get away from Longuet-Higgins, but the same technique cannot be used by a computer to get away from me. This is the force of your question really.
Well, I am saying that you are under the same difficulty as a computer. The sentence on the blackboard is a translation into question form, if you like, of the Gödel sentence to which we give a metamathematical interpretation although it is in fact a statement about numbers. It is the fact that you give the Gödel sentence an interpretation which makes it a question which the machine can't answer. It is the fact that I give this question an interpretation, or you give it an interpretation, that makes it a question which you can't answer. But if nobody gave it an interpretation then there would be no question which the machine couldn't answer and so the unanswerability of such questions depends upon them being recognised as questions. But in that respect you are in no better a state than a computer.
Now can I come back and ask you another question which I would like to raise. I think your argument, although you did not say so in quite so many words, rather takes it for granted that we can always see that there is this Gödel sentence and that it is true. Now this depends, I am assured by my better informed colleagues, upon seeing that the axioms are consistent. Now I think it is quite clear that given any arithmetic system we can't necessarily see that its axioms are consistent. For example, we don't know whether the ordinary axioms of arithmetic, with Fermat's theorem conjoined to them, are consistent or not. And it is simply not true to say that human beings are always in a position to find and to see the truth of a Gödel sentence.
Can I just first of all continue the first part of the argument. That is about this question, because I feel I do need to answer Longuet-Higgins slightly more fully. The crucial point is one which I have to feed in rather carefully, and that is why I brought in that awkward business about the smallest infinity that there is. The way that we can get round the difficulty of the question, namely is it rational to be able to give an affirmative answer to the question, namely… and then going round in a circle is because we can code both the questions and the algorithms of the computer onto the natural numbers.
But I have coded this onto the blackboard and the blackboard is the address of this question. What is wrong with having a blackboard as an address. Why do we have to have one of the natural numbers as an address?
Essentially because on the blackboard you would never get to the end of your sentence if you were to spell it out fully.
On the contrary, you can read it right to the end, and so can I.
Yes, but it fails in its reference.
It doesn't. It's perfectly clear.
I think I'll come in on a very different tack. You made some contrasts between the humanities and the sciences. The humanities deal with questions which are eternal questions, to which you can never give a clear-cut answer. For the sciences, you quoted Aristotle or Plato—that you can always tell whether a thing is made of lead or of silver, and you can settle this quite definitely and decide it; and that you gave as your paradigm of science. I think this is unfair—scientific questions are as eternal as those in the humanities are. We are still debating about the nature of the physical world. What happens is that for a time we have a theory about it, and during that time a whole lot of very dull science goes on, using that particular system—measuring this and measuring that, and what have you. But then after a time some absolutely new idea eventually boils up somewhere, a novel idea like relativity which changes the whole idea of material particles and their interactions. These are complete changes of scene, and changes of a whole set of algorithms. During a stable period, when people are all operating within the terms of Newtonian physics, you might say that you could program computers to carry out all these calculations by a set of algorithms. But if you suddenly had to make them work in terms of quantum mechanics you'd have to write a new set of programs. There is a great deal of humanities which is boring as a lot of sciences and that is saying quite a lot. You go away and read a lot of eighteenth century sermons, I don't think you will necessarily conclude that the whole of humanities deals with really intellectually exciting material. The point I want to make, referring back to this discussion of algorithms and computers that I am not a bit expert on, is that it may well be true, as you suggest, that there is always an incompleteness in any given system of algorithms, and I think there is always incompleteness in any given scientific theory. What has to happen then is that a new insight is brought forward and gets embodied in a new, more comprehensive set of algorithms, and you get a new type of scientific theory being employed. Where does this novelty come from? You spoke of it as a creative product of the autonomous individual, and that's one way of putting it, of course. Some people, possibly Freud, have tried to explain this creativity by references to internal mental events. These explanations don't by any means convince everybody, but I think one ought to try to go further than merely saying that a human being has an autonomous creativity. O.K., I think he has, but can we say anything about it?
I concur with most of what Waddington says. My account of the Senate of another university was not meant to be running down the sciences, or for that matter running down the humanities, but merely to bring up a point about different ways of arguing, different ideas of what constitutes a good reason or what constitutes an explanation. It is true, exactly as you said, that one keeps on having new insights in different sorts of sciences—then one formalises it, and then one produces an algorithm sometimes, although not as often as people think, and this is the way that science develops. All I have been trying to do is to give an a priori reason for saying that this must be so, and that anyone who has an a priori argument trying to fit all the sciences into a certain strait-jacket can be seen ab initio to be wrong. I think I otherwise agree with you.
I don't want to pick a quarrel with John Lucas about algorithms. He lined me up at the beginning of his paper as a great believer in algorithms because I said last night that if something was a logical law it admitted of no exceptions. I continue to believe that if something is a logical law it admits of no exceptions and I am very surprised that John should wish to deny this. Perhaps he does not. What I think he attributed to me last night was saying that if an argument worked in one case it must work in all parallel cases. That also I believe to be true, but of course the discovery of an argument for something is different from the discovery of an algorithm, and unfortunately we don't have algorithms for the discovery of arguments, although we may have for the testing of arguments. In some cases we do and sometimes not.
I would like to point to a limitation on the conclusion of John Lucas's arguments against the autonomy of computers which may not have been apparent. It is essential to his notion of a computer, and indeed, of course to any very strict notion of a computer, that it should work algorithmically. But I think nothing follows from what he said about the possibility of having an autonomous agent that worked electronically. The plain man's idea of a computer is not by John's argument ruled out from someday having the possibility of engaging in dialogue with human beings. Indeed I think that we have no absolutely good reason to believe that it might not be possible to have artifacts which were rational beings. It is only by testimony and induction that we know that we are ourselves not artifacts. It seems to me that though it would be very unlikely there would be nothing absolutely inconceivable in the idea that there might be a knock on my study door one day by a man saying ‘I've come from IBM to service you’, and then he opens me up inside and shows me all sorts of valves and things that I had no idea were there.
The second point that I want to make about John Lucas’ paper concerns the relation between the first and second part of his paper. The first part, you will remember, was devoted principally to logical considerations and the second part to moral considerations. Now I agree with John that Kant made too sharp a distinction between logic and moral philosophy. I think that there isn't an absolute distinction between other areas of philosophy and moral philosophy. I think that results in logic can be relevant to moral philosophy but I don't think that the results of logic which John mentioned were relevant to his moral conclusions. He expressed among other things a dislike of pharisaism. He may be right in his dislike of Pharisaism but I don't see how Gödel's theorem or Tarski's theorem has got anything to do with pharisaism unless he suggests that the reason why the pharisees of Jesus’ time behaved as they did was that secretly they were computers working on algorithms. I don't think that one can use Church's theorem to lead one into the church quite so quickly.
Can I start by apologising to Kenny that I maligned him about the laws of logic. I think if there is a difference between us it is infinitesimal, and I am glad to welcome him on my side. On the second point that he has made and the main point of Longuet-Higgins I feel some hesitation. I want to argue here but it is something that I have already done in print. I think I shall just allow myself a few words briefly on the possibility of a person being made of electronic hardware. The argument that I have been putting forward is entirely a logical one and has got nothing to do with the hardware. And Kenny is absolutely right to say that my arguments won't tell me what that animal is—that I meet on Forrest Hill—is made of. He might be made of DNA or he might be made of selenium cells. I can't tell. All I can tell is the principles according to which he is made, and although the IBM man might come and open me up, what he can't do is to service me in the ordinary standard sence because this would mean that he was going to put in the right algorithms, he knew exactly what I should be doing, and this is what the incompleteness arguments will tell against. That is to say, although it is a matter of empirical fact true that all the embodiments of minds that we meet on the face of the earth are born of women and begotten in the normal course of events, this is not a conceptual truth and it could be that there was as it were a very complicated thing with lots of wheels and wires which got beyond the capacity of any man to control, not simply as a matter of complexity but rather that it started having, as we might say, a mind of its own. That is to say, it decided things neither according to a random principle nor according to anything we could programme into it but which we could nevertheless recognise as rational.
I would like to take you up on that because in your book you have in fact argued from the nature of mind to some physical conclusions about determinism, in fact your argument essentially says that because of the nature of mental activity therefore human beings cannot be automata. I reject this assertion. But it seems to me that if you met somebody in Forrest Hill and they seemed to you to be rational you would have to conclude that they were not made of the usual computer hardware because at least ideally, though not in practice, computer hardware is deterministic in its operation. It never of course really is, because things go wrong; but in principle.
I go on the principle of it. That if it is deterministic in its operation that I can get a foothold for my argument; if it does operate algorithmically then the argument will go through. It is true, what Longuet-Higgins pointed out, that in order to establish the truth of Gödel's theorem one needs to assume the consistency of elementary number theory, but again I am not going to regurgitate this now. I think this can be shown to be a proper premise to assume, but it takes several pages of argument to establish that. Can I just take up a point that Waddington made, who through Freud.…
May I just say something first, because I'd like to put it in another way. It seems to me you are taking as your definition of rationality, as the criterion by which you decide if the thing you meet is a rational being, that it should at some point behave irrationally—that is to say, not following an algorithm. Is not this somewhat paradoxical? What is the distinction you are making between rationality and irrationality?
Well, I certainly deny that rationality is to be equated with following an algorithm. Here I would go over to a Kennyite position—but I shan't maintain it for the moment. I think Freud is a more popular topic. You see, one of the reasons why this century has seen a loss of faith in reason has been the work of Freud. Freud has forced us to recognise that our reasons are not at all what we thought them to be, and this has been as damaging for our self-respect as Darwin's showing that our ancestors were not what we had hoped they were. But the argument won't work. Those parts of Freud's work where he puts forward, as it were, quasi-scientific theories—he talks about surges and pressures and quantities of libido—has got very little value; and Freud as a straightforward scientist isn't the Freud you know of. The Freud you know of is the Freud of the novels and the interpretation of dreams, who is able to give extraordinarily convincing explanations of why people did things. Although on the surface they might have thought they were doing something for the best of motives, after Freud had heard one or two of their dreams and asked a few questions, he was able to reveal some other reason for their action—one which they were not conscious of. But if he was to be successful as a therapist his reasons in the end had to be ones which the patient could recognize as being his real reasons; and so the reason why Freud does not show us not to be capable of reasoning is because he is presupposing the standard form of understanding what are the reasons for actions. It is only because he is presupposing this, that he is then able to show that quite often we have been misleading ourselves about what our real reasons for actions are. That is, he is appealing from one level on the surface of the mind, in which we have one set of reasons, to another level of the mind, where we have another set of reasons; but those other reasons are of the same logical type. They are reasons for action, reasons which we can understand ourselves wanting to act upon, reasons which perhaps are not very reputable ones, but nevertheless rational ones, and not in the least bit the causal concomitances for the purely algorithmic exercises which is the alternative which is being put forward to us. Therefore I go along with Freud. I walk along to the end of the road, and then show that he is in fact an unwilling ally of mind, and not the antagonist that you had supposed.
Could I say something about the relationship between being deterministic and being algorithmic? I was arguing last night that something could be deterministic and yet have the properties of mind such as autonomy and freedom of the will, and tonight I wanted to agree with John that something which operated purely algorithmically could not have the properties of mind. But he said that if it works deterministically then it works algorithmically, and I don't see that this follows at all. There could very well be a deterministic system which operated deterministically, that is to say, each state of it was caused by the previous state, and no other alternatives were open, and yet there be no regular procedures or systematic correlations which would amount to an algorithm.
Here I think I almost have to invoke David Hume. A cause which links an antecedent state of the universe with a subsequent state of the universe, but is not a regular one, is one that I find very difficult to understand, if I'm approaching this from the standpoint of the scientist. I think the difficulty may be (and anyhow it's worth bringing out) that the word ‘determine’ is used in a wide range of senses, and the word ‘determinism’ often suffers from a certain ambiguity too. That is, ex post facto I can very often say why I did something, and I can often give the reasons why I did in terms of some antecedent state of affairs. The reason why I am here now I can explain, and I can explain fairly fully, but that sort of explanation, although it certainly explains and in that sense also can be said to be a cause of my being here, isn't the sort of sense in which we feel that something is carrying us along willy nilly or that the stars in their courses ordained that I shall be here, quite apart from anything that I decided to do. And so I think we need to press the word ‘determine’ rather carefully. I am using the word ‘determine’, and ‘determinism’ in a sense which can be explained in either a Humean or at least in some scientific sense, of some function which will correlate in a regular and universalisable fashion antecedent states of the universe with subsequent states of the universe. This is the sort of determinism which seems to me to preclude freewill. The fact that Tony kindly predicts that I won't start beating my wife tonight is one which I thank him for, but I don't in the least bit feel that it's a threat to my free and responsible action in refraining from it.