Y. Kageyama (Jan. 17)
Is critical rationalism more concerned with explanation?
I have read several chapters of your book, and the paper 'Truth, Rationality, and the Situation'. I am very interested in this paper, for I have translated chapter 8 of the Myth of the Framework.
Your remark that critical rationalism is more concerned with explanation, whereas instrumentalism is concerned with prediction has given rise to some questions in me, for I think that falsifiability, which is one of the main thesis of critical rationalism, is closely related to prediction. Test is always directed, or refers, to some future events, so that it is, in a sense, nothing but prediction. A mere explanation may be neither falsifiable nor testable, while any prediction is in principle testable, however vague and ambiguous it may be. This is, I think, the reason why a prophet has to dodge censure, when his prophecy does not come true. It is comparatively easy for us to see whether a prediction is true or not, while it is sometimes very difficult to see the truth of an explanation. Censure to false prediction or prophecy is often dodged by explanation. Thus I think that critical rationalism also is strongly interested in prediction.
M.A. Notturno (Jan. 17)
Prediction as a way to test explanatory theories
With regard to your question: falsifiability does have a lot to do with prediction -- I do not deny that. But Popper thought that we are concerned with prediction primarily as a way to test our explanatory theories. You say that 'Test is always directed, or refers, to some future events, so that it is, in a sense, nothing but prediction'. This is actually not true. The first way of testing an explanatory theory is purely internal. If it is self-contradictory, then it fails the first test. But a theory also fails if it retrodicts incorrectly. Indeed, the situation with prediction and retrodiction is, from a logical point of view, identical. There are, however, two reasons why we focus so strongly on prediction. The first is that we generally think that we already know the past (which of course is not always true), and we construct our theories so as to explain what we think we know. If a theory can't even explain what we think has happened, then we typically don't even worry about testing it against future events. The second is that we feel that theory must have something to it if succeeds in predicting something that has not yet occurred. This, however, is more a psychological than a logical point. We didn't know that there would be a solar eclipse on such and such a date, but this theory has predicted it, and now it's come true. The problem, of course, is that we can be overimpressed with such successful predictions.
You also say that 'A mere explanation may be neither falsifiable nor testable'. True. But if an explanatory theory is not falsifiable, then it is not scientific on Popper's view. End of story. I think that it is false, however, to say that 'any prediction is in principle testable, however vague and ambiguous it may be'. 'God will eventually send a plague to punish the non-believers' is a prediction. But how would you go about testing it? If a prophet is foolish enough to put a date on his prediction, then he may well have to dodge censure. But both scientists and prophets make predictions that turn out true. The difference between them is not their verifiability but their falsifiability. And what turns a prophet into a scientist -- though perhaps a scientist with false predictions (and hence false theories) -- is his willingness to make his prediction explicit enough for us to test it. But quite aside from that, I am not so sure that it is so easy enough to see whether a prediction is true or not. I have, in fact, been reading a new biography of Einstein. This led me to do a little research on Eddington's attempts to verify his predictions regarding the bending of the light from distant stars as it passes near the sun. I have now read several accounts of it, including Eddington's own, and it is interesting to see how the 'evidence' could be interpreted to corroborate Newton's theory instead of Einstein's. One point however is clear -- and Popper himself told me this and emphasized it to me when he did: Nobody would have been photographing the sky during the 1919 eclipse, and six months before and after, were they not trying to test whether Einstein's explanation was better than Newton's. I am not, however, denying the importance of predictions as tests of theories. But I do think that instrumentalists have a very different attitude. They are concerned primarily with prediction, and are willing to sacrifice explanation so long as their theories predict well. This I think is why Popper and Einstein were so opposed to the Copenhagen Interpretation of quantum theory.
Y. Kageyama (Jan. 18)
The advantage of critical rationalism
I have said: "any prediction is in principle testable, however vague and ambiguous it may be", because I don't think that, apart from prophecy, any statement with no reference to future date, however vague it may be, is a prediction. Accordingly, I would like to regard your example: "God will eventually send a plague to punish the non-believers" not as a prediction, but rather as what Popper called the pure existential statement, for this statement can be applicable to the past as well, without altering its central meaning. A prediction, I think, has some future period where its status might become in question; that is the predicted date.
I believe that one of the strong advantages of critical rationalism is its pragmatic value, which can be derived from the fact that it is interested in not only explanation, but also prediction. Most of the philosophical theories are only concerned with explanation. What is wrong with instrumentalism is, I think, that any instrument which does not correspond to reality turns out to be useless in the end.
M.A. Notturno (Jan. 18)
Instrumentalism and prediction
I don't see why reference to a future date should be a necessary condition for predictions. And if you stipulate that they are, then the sort of distinction that Popper wants to draw between falsifiable and unfalsifiable predictions becomes problematic. You say '"any prediction is in principle testable, however vague and ambiguous it may be", because I don't think that, apart from prophecy, any statement with no reference to future date, however vague it may be, is a prediction'. But the 'apart from prophecy' part is the whole thing. Prophecies are predictions. You may reply 'But they are not scientific predictions'. To which I reply 'Of course not, they are too vague and ambiguous for that'. I really do not see why "God will eventually send a plague to punish the non-believers" should not be regarded as a prediction. It is exactly like the sorts of predictions -- 'The revolution will someday come', 'Every young boy will have an Oedipus complex' -- that Popper criticized as unfalsifiable.
You see, my point in my last email is that there is no logical difference between predictions and retrodictions. Both are logical consequences of a theory and its initial conditions. I agree with you that one of the strong advantages of critical rationalism is its pragmatic value, and also that it is interested not only in explanation, but also in prediction. You say that 'Most of the philosophical theories are only concerned with explanation.' And perhaps this is true. But my own sense is that philosophers of science these days are more interested in prediction than in explanation. Popper's view, in any event, is that we are interested in prediction because we are interested in explanation, and because predictions provide the best opportunities for testing our explanatory theories. You say that 'What is wrong with instrumentalism is, I think, that any instrument which does not correspond to reality turns out to be useless in the end.' Perhaps. But the end may be a long time coming. What Popper was concerned with was that some instrumentalists, such as Bohr and Heisenberg, thought that the only thing that mattered was prediction. They were, as a result, willing to accept their 'Copenhagen' interpretation, despite the fact that they acknowledged that we could not understand it. Einstein, on the other hand, felt that explanation was the primary goal of science. So he was not willing to accept the Copenhagen interpretation, despite the fact that he was impressed by its predictions, because he thought it did not offer an acceptable explanation. So the matter is not whether or not we regard prediction as important. The matter is whether or not we regard prediction as primary.
Y. Kageyama (Jan. 19)
A thought experiment: an imaginary psychoanalyst
I admit I am somehow hasty in my phraseology. I apologize you, for by saying: 'apart from prophecy', I might have tried to rub out my inappropriate phraseology. And I also admit that I go too far in saying "however vague and ambiguous it may be". Yet I still hold my basic point.
One of my favorite stories of Popper's philosophy of science is the story he told in his 'Science: Conjectures and Refutations'. In this report, I have been strongly impressed by the contrast between the omnipotent explanation of psychoanalysis and the riskful prediction of Einstein's theory of relativity. Those who always try to evade any criticism against them usually scarcely offer prediction, but only explanation. So the psychoanalyst in Popper's report explains the behaviour "of a man who pushes a child into the water with the intention of drowning it; and that of a man who sacrifices his life in an attempt to save the child."(Conjectures and Refutations, p.35.) If I ask the psychoanalyst about another man in a similar situation: "what will the man do next? Will, or will not, he save the child?", I imagine the psychoanalyst would not reply to my question, or at best: "it depends on the situation". This is because to make prediction is always accompanied with some risk: the risk of turning out not to be true. If the psychoanalyst says to me: "even if the man next save the child, he will eventually push it into the water someday", I don't think this statement as a prediction, but as a mere evasion. In this sense, too vague prediction is, I think, not a prediction (I shall withdraw my phrase: "however vague and ambiguous it may be", and admit my mistake in phraseology).
I think the fundamental attitude of critical rationalism toward prediction is more serious than that of instrumentalism, because, as you say, "predictions provide the best opportunities for testing our explanatory theories", and because critical rationalists are more concerned with the truth of theories than with their usefulness or with persuading other people. In this sense, I believe that the concern of critical rationalism with prediction should be emphasized.
As to the instrumentalism in quantum mechanics, I agree with you in the following passages: "Einstein, on the other hand, felt that explanation was the primary goal of science. So he was not willing to accept the Copenhagen interpretation, despite the fact that he was impressed by its predictions, because he thought it did not offer an acceptable explanation." Yet I think that the main target of Popper's criticism of Copenhagen interpretation is not so much its instrumentalism as its subjectivism. If it were not for subjective elements in Copenhagen interpretation, Popper's criticism would not have been so incisive. One form of instrumentalism without subjectivism is, say, conventionalism, and Popper's criticism against it is, in my view, more modest than any other criticism against subjectivistic theories in physics.
M.A. Notturno (Jan. 20)
Being true and being testable are not the same
You say 'I admit I am somehow hasty in my phraseology. I apologize you, for by saying: 'apart from prophecy', I might have tried to rub out my inappropriate phraseology. And I also admit that I go too far in saying "however vague and ambiguous it may be". Yet I still hold my basic point.'
But your basic point, if I understood you right, is that every prediction is testable.
I agree with what you say about the Einstein story -- though Einstein seems to have been somewhat more sure about relativity (both special and general) than Popper pretends. I also agree with what you say about those who offer explanation and not prediction (compare this with what I say about political scientists in my open society paper). But what you say about the psychoanalyst confuses things a bit. You say 'to make prediction is always accompanied with some risk: the risk of turning out not to be true. If the psychoanalyst says to me: "even if the man next save the child, he will eventually push it into the water someday"'. But surely what the psychoanalyst says does run the risk of turning out not to be true. Being true or false and being testable are two completely different things. You make this point very well in the opposite direction in your paper: instruments are testable despite the fact that they are neither true nor false (and hence neither refutable nor falsifiable), though I do not think that Popper ever thought for a second that they were not testable, as you suggest. But a statement may be true and not testable. This is one of the critical differences between Popper and the positivists. It is why Popper was so opposed to psychologism, where being true and being believed (or justified, or testable) are one and the same thing. It is also why metaphysics, on Popper's view, is not nonsense. Popper thought that there is a lot of truth in Freud's theory. But whether or not it is testable is another story.
You say 'I don't think this statement as a prediction, but as a mere evasion. In this sense, too vague prediction is, I think, not a prediction (I shall withdraw my phrase: "however vague and ambiguous it may be", and admit my mistake in phraseology).' But this is really an evasion. The psychoanalyst makes what he regards as a prediction. And he does seem to be saying what will happen in the future -- though he is certainly not specifying a date. You call it an evasion and criticize it for not specifying a time frame. OK. But why should that mean that it is not a prediction? Why can it not be both an evasion and a prediction? The only reason I see is to save your claim that any prediction, no matter how vague or ambiguous, is testable. When we find what would otherwise be regarded as an untestable prediction, we say that it is not a prediction (presumably because it is untestable!).
You say 'I think the fundamental attitude of critical rationalism toward prediction is more serious than that of instrumentalism, because, as you say, "predictions provide the best opportunities for testing our explanatory theories", and because critical rationalists are more concerned with the truth of theories than with their usefulness or with persuading other people. In this sense, I believe that the concern of critical rationalism with prediction should be emphasized.' I agree. I would go further and say that the instrumentalists' concern with prediction is really a concern with power, and with controlling nature, instead of understanding it. I do not want to underestimate the role of prediction. But I do not want to overestimate it either. And I do want to keep the priorities straight. Popper thought that we are concerned with prediction because we are concerned with testing our explanations. The instrumentalists are concerned with prediction for its own sake.
I certainly would not deny that Popper was opposed to the Copenhagen interpretation's subjectivism. But take a look at his introduction to the quantum theory volume in the postscript. He explained it to me once in the following way. Einstein had made a breakthrough in physics by utilizing what were, for all intents and purposes, some instrumentalist philosophical insights. The younger physicists -- Bohr, and still (much) younger Heisenberg -- were very impressed by this, and they somehow inferred that in order to make a greater breakthrough they had to introduce a still more radical instrumentalism. So they were quite content to forego the whole issue of whether the theory gave an explanation and whether it could be understood in order to do so. They even went so far as to interpret the idea that it could not be understood as signifying its correctness. I sometime (glibly) give a psychological explanation for this. Heisenberg, when he came to Bohr, was about twenty years old. When someone is twenty years old, he may be quite ready to give up reality for a theory.
You are right that Popper's criticism against conventionalism is more modest than his criticism against subjectivism. But this is because he thinks that conventionalism is a defendable position. But he also thinks that it is a false position.
Y. Kageyama (Jan. 20)
Testability means predictability
Our discussion on what is to be called a prediction seems to have become a dispute of the meaning, or definition, of 'prediction'; but I don't want to dispute about the meaning of word, so I shall concede you when you say: "... why should that mean that it is not a prediction? Why can it not be both an evasion and a prediction? The only reason I see is to save your claim that any prediction, no matter how vague or ambiguous, is testable."
You say: "...your basic point, if I understood you right, is that every prediction is testable". But, rather than this, I would like to say that every test is somehow a prediction, for every serious test must specify what will be the result before the test will be carried out.(This idea leads to the principle of accountability, argued in The Open Universe: An Argument for Indeterminism. In my view, falsifiability and accountability are closely related to each other.)
I was ambiguous about whether prediction is equivalent to test or not. Yet your criticism of my idea made it clear that every test is a kind of prediction, but not vice versa; too vague prediction may not be testable.
Still, the reason why I stick to prediction is that every test is a kind of prediction: testability means predictability (but not vice versa).
When I first read in your paper, 'Truth, Rationality, and the Situation', the following passage: '...the problems that instrumentalists want to solve are primarily problems of prediction, and the problems that critical rationalists want to solve are problems of explanation.', I felt that the significance of prediction was underestimated. Not only critical rationalists, but also astrologers, psychoanalysts, or even metaphysicians such as Hegel are interested in explaining human behaviours or the structure of the world.
I want to refer here to the reason why Popper questioned the way Freud put his theory, even though Popper admitted that Freud's theory may contain some truth. It is, of cause, because it was not testable. And as I said before that testability means predictability, so if a theory has no power of prediction, it is not testable; predictability is a necessary (not sufficient) condition for testability. This is the reason why I regard prediction as important.
In my view, a prediction may be somehow seen as an explanation of the unknown, and in this sense, I agree with your remark that 'the problems that critical rationalists want to solve are problems of explanation', for, as I have said elsewhere, that falsifiability means the openness to the unknown.
M.A. Notturno (Jan. 22)
Self-consistency and retrodiction
We are, of course, talking about the meaning of 'prediction'. But I don't think that the dispute is merely about words. You say that 'every test is somehow a prediction, for every serious test must specify what will be the result before the test will be carried out'. You want to link predictability with testability, and this led you to say that every prediction must specify a date. My example of vague and/or ambiguous prophecies has led you to concede that there may be predictions that are not testable. But you still want to say that 'every test is a kind of prediction: testability means predictability (but not vice versa)', and I do not see why even this is true. A test of a theory is a sort of hurdle that it must leap in order to be allowed to continue -- in order not to be eliminated. But the first such hurdle that a theory must leap is self-consistency. If a theory is self-contradictory, then it does not, on Popper's view, pass the test. But where is the prediction here? You may say that there is a prediction that the theory is self-consistent, or that it is not. But I really don't see why we need predict this at all -- we may simply test for it. And it is not, in any event, the sort of thing that we usually mean by 'something that the theory predicts'. Or consider retrodictions. Suppose, for example, that you had a theory of planetary motion that got the dates for eclipses in the 19th century wrong. I think that we would say that the theory failed a very important test -- the test, namely, of accounting for all the known data. This is indeed a test that theories do fail -- though some theories, like Newton's and Einstein's, actually end up correcting the known data instead! But where, again, is the prediction in this test? You may, again, say that there is a prediction that the theory will account for the known data. But that, again, is not what is usually meant, and my sense is that the only reason anyone would argue for it is to preserve the idea that every test involves a prediction. My sense is that you may be using 'prediction' to mean 'entailment'. We test a theory by the statements that it -- along with initial conditions, background knowledge, etc., etc. -- entails. This is because a statement cannot entail a false statement without itself being false. If a statement entails obvious falsehoods, such as contradictions, then we regard it as false. But not all of the statements that a theory entails are about the future. And they are thus not all predictions.
You say that 'When I first read in your paper, 'Truth, Rationality, and the Situation', the following passage: '...the problems that instrumentalists want to solve are primarily problems of prediction, and the problems that critical rationalists want to solve are problems of explanation.', I felt that the significance of prediction was underestimated. Not only critical rationalists, but also astrologers, psychoanalysts, or even metaphysicians such as Hegel are interested in explaining human behaviours or the structure of the world.'
Yes, most everyone -- except, perhaps, instrumentalists and authoritarians -- are interested in explanation. And of course, what distinguishes critical rationalists from the others is that we advocate testing our explanations. And of course, one way of testing our explanations is by checking whether or not the statements that they predict are true. I do not deny any of this. Nor do I underestimate the importance of prediction. My point, rather, is that critical rationalists are interested primarily in explanations, that they are interested in predictions as a way of testing their explanations, and that they are not, for that reason, satisfied with theories that, like the Copenhagen Interpretation, give good predictions and bad explanations.
If Freud's theory is not testable -- and Popper, I think mistakenly, acknowledged, under the influence of Bartley, that parts of it were (see my paper on psychoanalytic theory) -- it is not because Freud made no predictions. On the contrary, he predicted that all newborn male human babies will develop Oedipus complexes. But he also retrodicted that all dead human males had Oedipus complexes. The untestability, on the contrary, was due to the difficulties of determining, independently of the theory, whether and when someone has an Oedipus complex.
You say 'In my view, a prediction may be somehow seen as an explanation of the unknown, and in this sense, I agree with your remark that 'the problems that critical rationalists want to solve are problems of explanation', for, as I have said elsewhere, that falsifiability means the openness to the unknown.' I agree that falsifiability involves an openness to the unknown. But in my view, predictions are not explanations, but things to be explained -- hopefully by the theory that predicts them.
Y. Kageyama (Jan. 25)
The notorious ambiguity of the concept of understanding
I have been looking forward to your reply especially this time, for I am not so confident of my thesis: "every test is a kind of prediction: testability means predictability". In order to learn from you, I have expected it to be criticized by you. So I have dared to advance my uncertain thesis as a bold conjecture.
You say 'A test of a theory is a sort of hurdle that it must leap in order to be allowed to continue -- in order not to be eliminated. But the first such hurdle that a theory must leap is self-consistency.' I completely agree to this. But in saying 'every test is a kind of prediction', I mean only empirical test. My usage of the term 'test' in this way can be permitted, since Popper's falsifiability is only concerned with empirical theories.
Yet your second objection using the cases of retrodiction is really a blow to me. I have never thought of the cases of retrodiction. If I still insist on that retrodiction is a kind of prediction, I would come off the common usage of the term 'prediction'. I don't want to defend my thesis for its own sake, so that I must admit I am in a stalemate in this point.
Yet, through our discussion, I come to think that we should say that critical rationalism is primarily interested in test rather than in explanation.
I still want to keep my vigilance concerning the concept of explanation, because explaining something is for the sake of understanding it and the concept of understanding is, compared with prediction, notoriously ambiguous. Some advocators of the Copenhagen Interpretation would say that they are presenting the way to understand nature, so a kind of explanation in their own way. As Popper says, quantum mechanics "provided the (probabilistic) theory of what may be called their 'animation'." (The Myth of the Framework, pp.164-165.)
M.A. Notturno (Jan. 27)
More on retrodiction
We agree about the self-consistency test. It is not an empirical test. But you do need to consider retrodictions more seriously. The most important 'corroboration' of Einstein's theory, for example, is that it explained the perihelion of Mercury. This was far more clear than the corroboration of the measures it predicted for the bending of light around the sun, or the corroboration of its predictions regarding the redshift. All of these statements -- about the perihelion of Mercury, the bending of light, and the redshift -- followed from his theory given certain initial conditions, etc., etc. We can say that Einstein's theory 'predicted' the perihelion of Mercury. But all we really mean is that it was a logical consequence of the theory. We already knew about the perihelion of Mercury. We also knew that it posed a problem for Newton's theory. People had tried unsuccessfully to solve the problem by 'predicting' the discovery of a new planet closer to the sun. But what makes the one a retrodiction and the other a prediction in these cases seems to have more to do with our knowledge than with anything else. I mean, if light bends around the sun according to a certain measure, then presumably it has always bent around the sun according to that measure. We say that the theory predicts it, partly because it follows from the theory, and partly because we didn't know the fact before we tested for it. This is why I want to emphasize that we can easily be overimpressed by successful predictions. Logically speaking, they stand in the same relation as retrodictions. But they are psychologically more impressive, precisely because they speak of things that we do not already know. You still want to say that critical rationalism is primarily interested in test. I do not want to deny the importance of testing. Certainly what separates critical rationalists from many others is the emphasis that they put on testing. But why are we interested in tests? What are we testing? Popper's answer to this is loud and clear. We want to explain, and to test our explanations! This is why Popper was dissatisfied with the Copenhagen theory. Its predictions were wonderful. But it didn't give an intelligible explanation for them.
Y. Kageyama (Jan. 29)
The reason why I thought testability means predictability
Your email on 27th is highly instructive for me. I have pondered over the reason why I dared to advance such a feeble thesis: 'testability means predictability'. Just as you say, I have thought of a logical entailment, in asserting this thesis. Both of the statements to be tested and to be predicted are logically deduced from some premises which may be constructed by some universal theories and initial conditions. Though a logical entailment is, strictly speaking, timeless, its conclusion is, when we manipulate, always posterior to its premises. In this sense, test and prediction are of the same form. Yet, while a test is indeed timeless, a prediction always refers to the future, as opposed to retrodiction.
My understanding of the roles of prediction and explanation in critical rationalism has been greatly improved through our discussion.
Through our discussion, I also have had an exciting experience which would be similar to that of students in Popper's seminars, described by Bartley as follows (and you are referring to in your paper): "Popper interrupted every sentence; nothing passed unchallenged: every word was important." I must say thank you for this precious experience.
M.A. Notturno (Jan. 30)
Importance of prediction as a logical entailment
I think we have reached agreement here about testing and predictions. But I frankly do not think that it is really so much teaching as it taking another person seriously and respecting him. This is what I think teaching -- or rather learning -- is all about. Not that other thing, where people mindlessly repeat things that others have said and think that they must be true simply because they have repeated them. (p, p, p, p, p, therefore p!) It was, during our discussion, just as possible that I would have learned that testing necessarily involves prediction after all. But the argument went the other way, and partly because you were not willing to make arcane and question-begging distinctions just to save your thesis. That is why I said that I liked you. And your thesis, incidentally, was not so feeble. Predictions are very important for testing theories. My only point is that we can easily misunderstand why they are important, and thereby exaggerate their significance. They are important not so much because they speak about the future, but because they are logically entailed by a theory. This is what enables us to use them to test the theory. But they do not themselves entail the theory. And a theory's predictions may be true and the theory itself false -- which would, I take it, be impossible in an instrumentalist account. I want to stress this for two reasons. The first is that it shows the importance of deductive entailment for science -- something which is now seriously threatened, not only by overt inductivists, but also by people like Larry Lauden who now want to deny that evidence for or against a theory needs to be entailed by it. The second is that it shows that there is nothing logically special about the fact that a statement speaks about the future, and that we should not overexaggerate the importance of predictions by saying that they show that the theory must be true, or by making them an end in themselves. Many people seem to think that the fact that we were able to predict in great detail somehow provides supporting evidence for the theory -- as opposed, say, to supporting evidence for the theory relative to another that does not make the prediction (which really means evidence for eliminating the theory that does not make these predictions).
I am also wondering what our next philosophical discussion will be about.