This document discusses the goal of prediction in social science. It outlines three major views on the nature of social sciences - whether they are like natural sciences, whether causal explanations are possible, and whether social theory has social causes and effects. The document focuses on the view that a key goal of developed social science is large-scale social prediction. It argues that explanation and prediction are logically identical, as both rely on generalizations and deductive reasoning. Successful explanation implies an ability to predict future events. The document also discusses challenges around predicting social trends versus laws, and distinguishing logical from methodological considerations.
This document discusses the goal of prediction in social science. It outlines three major views on the nature of social sciences - whether they are like natural sciences, whether causal explanations are possible, and whether social theory has social causes and effects. The document focuses on the view that a key goal of developed social science is large-scale social prediction. It argues that explanation and prediction are logically identical, as both rely on generalizations and deductive reasoning. Successful explanation implies an ability to predict future events. The document also discusses challenges around predicting social trends versus laws, and distinguishing logical from methodological considerations.
Original Description:
The Philosophy of the Social Sciences by Alan Ryan
Original Title
The Philosophy of the Social Sciences by Alan Ryan-204-225
This document discusses the goal of prediction in social science. It outlines three major views on the nature of social sciences - whether they are like natural sciences, whether causal explanations are possible, and whether social theory has social causes and effects. The document focuses on the view that a key goal of developed social science is large-scale social prediction. It argues that explanation and prediction are logically identical, as both rely on generalizations and deductive reasoning. Successful explanation implies an ability to predict future events. The document also discusses challenges around predicting social trends versus laws, and distinguishing logical from methodological considerations.
This document discusses the goal of prediction in social science. It outlines three major views on the nature of social sciences - whether they are like natural sciences, whether causal explanations are possible, and whether social theory has social causes and effects. The document focuses on the view that a key goal of developed social science is large-scale social prediction. It argues that explanation and prediction are logically identical, as both rely on generalizations and deductive reasoning. Successful explanation implies an ability to predict future events. The document also discusses challenges around predicting social trends versus laws, and distinguishing logical from methodological considerations.
Up to this point, we have examined the two major and rival
theses regarding the nature of the social sciences that were outlined in the opening chapter of this book. We have seen that although there are some deeply disturbing anomalies in the claim that there are no important differences between the human sciences and the natural sciences, these anomalies do not eliminate inquiries into the causes and effects of social activities, and they do not much impinge on our initial account of satisfactory causal explanation. In these two last chapters, we shall try to justify the concerns of the third thesis we mentioned, that which stresses the social causes and effects of social theory. In this chapter we shall explore the implications of the long-standing view that the aim of developed social science is to produce the prediction of large-scale social changes, and that its maturity as science can be estimated by the yield of such predictions; in the final chapter we shall raise some of the problems of 'objectivity' in the social sciences, and briefly discuss the topic of ideology. What provides a common link between these chapters is the fact that the social sciences are preeminently 'policy sciences'; {1) that is, they have been developed by and for men who have wanted to use the knowledge they could gain to bring about changes of one or another kind. Equally importantly, they may well have policy effects, even where they are not developed for policy purposes - just as any other science may spawn technologically relevant side-effects, though research was not initiated for technological reasons. For in the case of the social sciences, one important con- sideration is that how people act in society depends on what they believe about society: if they come to believe a different story, they will also come to behave differently. This fact was as well-known by Burke, who as a conservative wished to limit social speculation, as by Marx, who as a revolutionary 197 wished to harness it to the cause of revolution. In this chapter then, we shall consider three issues. The first is the validity of the claim that explanation and prediction are logically identical, merely two faces of the same knowledge, with the apparent implication that success- ful explanation must entail an increased ability to predict future events. The second problem begins at this point, for it is that of the feasibility in principle of projects for large-scale sociological prediction; and the final question again stems from this problem, for - in the light of such current phenomena as committees set up to predict what will happen over the next thirty years - we must distinguish between law-based and trend-based predictions, and evaluate Professor Popper's claim that the neglect of this distinction has been fatal to much sociology. (2) In the course of discussing these issues, we shall be able to see rather further than before into the distinction between logical and methodological con- siderations, and also to see how these are - and are not - related to substantial claims about the real world. The claim that explanation and prediction are logically 'isomorphic' - that is, the claim that they are formally identical - rests essentially on a simple logical point, made earlier in our account of deductive explanation. According to the deductive view of explanation, as we then described it, explanations when completely spelled out are deductive arguments. To explain a particular occurrence, we have to be able to deduce the statement that this occurrence took place as and when it did from the general law covering that class of case together with some statement(s) about the holding of the initial conditions for the occurrence. In the terminology of causal connection, we assert the dependence of the event to be explained on its causal antecedents in the light of an underlying causal generalisation. Thus we could explain the way in which Thomas Jones voted in the British general election of 1966 by pointing to the fact that he had been a staunch supporter of the National Union of Mineworkers for the past thirty years. We know, now, that such an account is not straightforwardly an explanation, by the rigorous standards of such 'deductivists' as Professor Hempel: it is rather an 'explanation sketch', (3) the outlines of what could, were there any chance of people being misled, be trans· 198 formed into a 'complete' explanation. Moreover, we also know by now that to cite membership of the N.U.M. as a causal antecedent is subject to the qualifications we made about the role of reasons as the causes of human action - membership of the N.U.M. is a shorthand indication of the kind of reasons for voting which will move Thomas ] ones. Still, these are only qualifications; and the important point remains the deductive mould into which we can put such explanations. And in the case of this explanation, there is no difficulty about doing just that. With the backing generalisa- tion that 'All staunch, long-standing members of the N.U.M. vote Labour' and the initial condition that 'Thomas] ones is a staunch, long-standing member of the N.U.M.' we can infer that 'Thomas] ones votes Labour'. The key element in all this is the backing generalisation, for it is on this that the identity of explanation and prediction hangs. It will be recalled from our earlier discussion that a general statement of the form 'All X are Y' is not to be analysed as a conjunction of singular statements about all the Xs which happen also to be Ys, but is rather to be read as a hypothetical statement to the effect that 'If anything is an X, then it is a Y', and as such a statement it is said to be strictly speaking tenseless, and to have no particular spatia-temporal reference. Thus 'All long- standing, staunch members of the N.U.l\1. vote Labour' is to be read as 'If there is someone who is a staunch, long- standing member of the N.U.M., then he votes Labour'; and as we saw, the formulation of such statements in the predicate calculus as statements of the form (x) (Fx- Gx) is designed to do justice to the logical structure hidden beneath the grammatical surface. It is important to be clear about how we read a symbol such as ' - ', as 'if ... then'; for it is equally about past, present, future and hypothetical cases, so that 'All X are Y' is to be read not just as 'If anything is an X it is also a Y', but equally as 'If anything was an X, then it was also a Y', 'If anything is (in future) an X, it will be a Y', 'If anything were to be an X, then it would be a Y'. (4) This, it is plain, means that not merely are explanation and pre- diction logically isomorphic, but so equally are explanation and retrodiction, and explanation and hypothetical pre- diction. That is to say that the generalisation about the relationship between membership of the N.U.M. and voting 199 Labour equally governs such statements as 'Thomas Evans was a long-standing and staunch member of the N.U.M., so he must have voted Labour' or 'If Thomas Evans had been a staunch, long-standing member of the N.U.M., he would have voted Labour'. It goes almost without saying that this feature of explanation is noticeable in day to day life, where we are very frequently called on to take advantage of i i - especially where we are concerned to reproach someone. Parents are often to be heard telling their children 'If you had grown up in the kind of household we did, you would take more notice of what your father tells you', a counterfactual utterance clearly predicated on the assumption that the children have not grown up in such a house, and do not take much notice of what father says. Radicals reproach conservatives in such terms as 'If you were a Bolivian peasant, you would regard Che Guevara as a hero', which again is predicated exactly on the belief that the person addressed is neither a Bolivian peasant nor an admirer of Che Guevara. And, of course, pre- diction and explanation run everywhere hand in hand: parents are likely to tell their children 'If you bring up your children as permissively as we've brought you up, you'll be sorry'; and the conservative is likely to tell the radical 'When you reach my age, and know more about the world, you'll see it's not so easy to change things for the better'. Now, what all this means is that the only differences between explanation and prediction lie in whereabouts in the deductive argument we begin. With the case of explanation, we begin with the event to be explained, i.e. with the con- clusion of the argument, and look for the appropriate generalisations and initial conditions from which to deduce this conclusion. In the case of prediction, we begin with the general- isation and the initial conditions and then go on to forecast the coming occurrence. And in the case of our hypothetical inferences, we simply suppose the initial conditions, and then draw the inference as to the consequences of this supposition. And one illumination which this sheds on a contentious point is worth noting. The social sciences are often said to suffer drastically from their inability to engage in controlled experiment; but this charge is also often said to be nothing more than a practical inconvenience, shared with such strikingly successful sciences as astronomy, and thus not in any 200 sense a logical handicap. (5) The grounds for this charge and rebuttal are now clear. If we are concerned with the testing of explanations, and especially of the underlying generalisations on which they rest, experiment will be extremely useful, in that we can produce the initial conditions and see whether the conclusion holds. Obviously, such a process gives us a degree on control over science that is otherwise difficult to obtain. But it is not logically necessary, in that we can wait until nature provides our experimental situations for us, or we can infer with care from the situations we do observe in nature to those which we cannot observe or create, but can at least envisage. But this is a merely practical consideration. It is plain that as a matter of logic, the identity of explanation and predic- tion is unshakeable. (6) However, this by no means entails that the goal of science is to try to produce wide-ranging predictions in practice. Two considerations come into play at this point. The first concerns the kind of generalisations available. We saw earlier that valid deductive arguments required universal generalisations as their general premisses, and that where we have not got such laws, we often offer explanations as the best available, without claiming that they are conclusive. The second concerns the consequences of the hypothetical analysis of causal generalisations. A statement to the effect that 'If anything is an X it will also be a Y' does not tell us whether anything ever will be an X -- and hence cannot tell us whether anything will be a Y. A generalisation such as 'All trades unionists vote for radical parties', if it is really meant as a causal law - and we shouldn't forget that it may not be, that it may be no more than a summary of the behaviour of trades unionists at the present day - does not tell us anything about voters for radical parties in, say, 1983. Unless we have such singular statements as 'there will be at least 750,000 trades unionists in 1983', we have no predic- tion at all; in short, we need statements about initial condi- tions if we are to utilise hypothetical generalisation for positive predictions. Now, it is not a matter of logic whether we can in fact produce either appropriately universal generalisations or secured statements about initial conditions; it is a matter of how the world is actually constructed that determines whether such statements and such generalisations can be established. Methodological claims about what a 201 developed science ought to aim at thus depend both upon assumptions about the logical requirements of explanation and upon reasoned guesses as to what the world is like. What will not do - and what has tended to happen - is to project onto the world the image of strict deductive argument and pass that off as the methodological aspirations of social science, just as it will not do to project the present weakness of social science explanations back onto logic and claim that they invalidate deductive reasoning. Two questions dominate this issue. The first is whether the social sciences can produce anything other than approximate laws and generalisations, and produce genuinely universal laws with an adequate spatia-temporal range of applicability. The second is whether we can make the necessary singular predictions about initial conditions, such as would enable us to draw from our generalisations predictions of a positive kind. The two issues are connected in that predictive success and failure feed on themselves: if we possessed strictly universal, non- probabilistic generalisations, then we should be able to predict with accuracy the states of affairs which form the causal antecedents of those states of affairs which are the subject-matter of more distant prediction. And, equally, if we cannot predict with exactitude what will happen in the near future our guesses as to the farther future will become rapidly no more than guesses. Now, it has been argued by some writers that the supposed isomorphism between explanation and prediction defended above is mythical, since there are successful and developed branches of the physical sciences where we can explain in general what is going on, but cannot predict particular occurrences. Professor Hanson argued this for the case of quantum mechanics, (7) where we cannot predict single quantum jumps, and it could be extended to genetics, where we can tell after the event why a successful mutation survives - in terms of the theory of natural selection - but we cannot predict before the event why a particular mutant should appear. Now, the obvious retort to the claim that here we have explanation without the possibility of prediction is that this rests on an ambiguity about what is explained. In neither of these cases is there any claim that we can explain the single event which we cannot predict; that is, we can explain 202 the mutation's survival, but we cannot explain its initial appearance, and it is this, and this only, which we cannot predict. Or we can explain the probability of cases of quantum jumps, but we cannot explain why one jump occurs - and it is this that we cannot predict. In other words, what we cannot explain we cannot predict, and conversely. (8) Here as elsewhere, there is nothing to be gained by quarrel- ling over the application of the label 'explanation' to some particular theory. But what can be said is this: where we have the result - after careful and prolonged scientific investiga- tion - that no determinate laws are discovered, and we also have a general theory which explains why this should be so, then methodologically it is a futile occupation to bemoan their absence. In social science it is presently the case that we are quite well equipped, but with approximate and not very far-reaching generalisations about how people behave in given situations. Thus we have rather a good idea of what motivated voters in Britain during the 1950s, but the explanation of why the Conservative Party remained so consistently in power would be extremely hard to extra- polate to the United States of the 1950s, or to Britain in the 1960s, let alone to more remote societies and times. (9) Such generalisations thus offer us little in the way of backing for detailed prediction, though they may be invaluable in providing us with day to day help in assessing our situation. The other great problem for the social sciences is establishing before the event rather than after which causal antecedents are the important ones. This is obvious enough, in the light of the need for initial conditions in causal explanation. But there is a less obvious point relevant here. We have argued so far that explanations of human affairs rest, directly or indirectly, on what people have reasons for believing and intending under the situations they are in; and this is ineluctably to introduce into the initial conditions of social events the beliefs and feelings of the social actors about those initial conditions. Reasons are only effective reasons in the light of beliefs and attitudes; but beliefs and attitudes are also subject to more or less conscious reappraisal. Whether or not this is a vital difference in the logic of explanation of human affairs is momentarily beside the point; what is certainly true is that (in any situation other than one where there is complete 203 control over what people can believe) it is as a matter of fact much easier to reconstruct from hindsight what a person or a group must have believed and wanted than it is to predict this in advance of their acting at all. If the only evidence about the possibility of prediction were culled from our experience of everyday explanation and explanation in the social sciences, it is doubtful that anyone would have thought that scocial scientists ought to aim at establishing long-term predictions. The image which seems for so long to have dominated the intellectual scene, however, is that of astronomy with its predictive triumphs dating back so far as the Babylonians. There has indeed been a debate since antiquity couched almost in the same terms that Plato and Aristotle gave it; the former, impressed with astronomy, and believing that a uniform natural order must operate on uniform principles throughout, put forward astronomical theories about the cycles of change and decay and growth in social life; {10) Aristotle, judicious, impressed by the multiplicity of different natural phenomena, insisted that we ought only to aim at the kind of exactness which the particular subject-matter permitted, and that it was folly to erect anything grandiose on the weak foundations of our social knowledge. ( 11) But the attractions of a predictively successful social science are obvious enough. The belief that there is a 'pattern' in history is one of the most \\idespread of human beliefs; Stone Age tribes have their myths, and contemporary states their generally accepted accounts of their history which are often as mythical. There is not a great leap from believing that there has been a pattern in what has happened in the past to believing that this should provide evidence about what can be· expected to happen in the future. Of course there has usually been a good deal of ambiguity about whether the guide is to what will happen, or whether it is a guide to what ought to happen; but this ambiguity is grounded in our own doubts about how much of our future is under our control, and is as common in sociological theory as in earlier theological and mythological accounts of history. So obvious is this kinship between the theological systems evident in Judaism, in such works as St Augustine's 'City of God', (12) or in Plato's account of the 'Golden Year', that 204 more than one writer has argued that an allegedly sociological system ought to be analysed as in reality a theological system because of the needs it fulfils in this way. Thus Professor Tucker argues that Marx was a covertly reli-gious thinker, and that the sociological mask worn by his work is largely delusive. (13) But it is equally true that many of the founders of the contemporary sociological tradition happily accepted that what they were doing was, so far as its social-purpose went, a similar undertaking to that of the great religious systems of the past. Saint-Simon, for instance, proposed to found a New Christianity on Newton's law of universal gravitation, a cornerstone to replace the old Christian claim that God is love; and Saint-Simon's (14) ambitions rubbed off on his greatest disciple, Auguste Comte, who gave a good deal more than merely the word 'sociology' to the subject. Comte shared with many other nineteenth-century thinkers the belief that the sociological and cosmological theories associated with Christianity were played out, and that they no longer gave people the sense of where they had come from and where they were going to which all societies needed if they were not to be overtaken by anarchy. Sociology, and especially a sociological account of hist<;>ry, together with predictions about the way society would go on developing, was intended to fill the gap left by scepticism and secularisa- tion. Even those nineteenth-century thinkers to whom Comte's mature, and bizarre, attempts to reformulate medieval Catholicism and its hold on social life held no appeal at all, saw that he grasped a point of great importance: they agreed that there was a need for a scientifically respectable account of how modern society had come about, what its problems were, how it was solving them or failing to solve them, and what means were available for the improvement of the future it faced. (15) De Tocqueville, :tvlarx, Durkheim and Weber all faced up to this need in their different ways. It is, however, necessary at this point to reiterate what we have already said, that the aspiration to this kind of predic- tive social science was not an aspiration which anyone could have contracted from a consideration of its existing achieve- ments. Rather, it was culled from the apparent need for the kind of doctrines which traditionally religion had provided and was now failing to provide, and from an assumption, 205 largely fed by the success of the natural sciences, that the social universe was predictable in the same way as the rest of the natural world. The less a thinker felt this homogeneity between the social order and the remainder of nature, the less likely he was to aspire to prediction in sociology. Comte, Mill and Marx all emerge as more inclined in this direction than were, say, Weber or Durkheim. The assumption which dominates Mill's picture of the goals of social science is the Newtonian assumption that the universe is a determinist order; what this means is that it is in principle possible to predict every state of the universe from a consideration of any other state together with a full knowledge of the causal laws operating .at that time. What is true of the universe in gross is true in detail of any subsystem within the universe which i~ sufficiently isolated to be studied on its own. From a consideration of the system at any time, we ought to be able to predict (or retrodict) its state at any other time. Writers like Mill were explicit about the origins of their assumptions, and references to 'celestial mechanics' and to hopes of a 'mecanique morale' are frequent. {16) The socio- logical implication is that since societies are also determinist systems, a full knowledge of any society at a particular point, together with a knowledge of the causal laws operating on it, should allow us to predict its state at any other time. Of course, everyone agreed that the actual computation of these predictions was quite impossible. The aim was to gain agreement on the principles involved. However, it is apparent after a moment's thought that this is a much less plausible programme than it appears at first sight. In exploring the doubts we must have, we can answer the question posed in Chapter 1 as to whether the kind of determinism and indeterminism in social science is like that in the natural sciences, and also enforce the lesson we have already received about the need to distinguish between the demands of logic and the possibilities allowed us by the phenomena. The first problem we need to raise is the ambiguity we earlier touched on, as to whether religious, mythic or sbciological predictions tell us what will happen or what ought to happen. The problem is this. Ordinarily we have a practical motive for wishing to know what will happen next; we are not motivated by idle curiosity about what will 206 happen to our society. Rather, we want to be able to act: we want to ensure that favourable trends will continue, and that unfavourable trends will be stopped, or where this is impossible that we shall circumvent their ill-effects. But now we face the problem we raised in Chapter 1 about the inter- action of the prediction and the behaviour predicted. The initial puzzle is that if the prediction is meant as a prediction of what will happen no matter what we do, then it is no use as a guide to action. We can only utter an unconditional or categorical prediction - what Popper calls a 'prophecy' - when we are sure that no action of ours will materially affect the outcome. (17) Thus once the Wall Street crash of 1929 was well under way, we could have predicted that prices would slump all the way down, because we knew that in a free market for securities, a panic once it has begun cannot be terminated by the deliberate actions of buyers and sellers in that market. This is analogous to predicting of a man who has fallen from the top of the Eiffel Tower that he will fall to the bottom - once started, there is absolutely nothing he can do to stop himself. But unlike the case of a man falling from the Eiffel Tower, the Wall Street crash prediction is likely to be unrepeatable. For men learn from their mistakes, and thus come to modify their future behaviour. Thus once it emerges that the predictability of the Wall Street crash depends on the fact that a freely operating stock market of the 1929 variety does not allow for co-operative efforts to slow down or avert disasters, the natural result is a modification of the system, and hence a falsification of attempts to repeat the success of the previous prediction; and this, of course, is a major point of difference between the kind of events which form what we term the Wall Street crash and those which form the sequence of someone's falling from the Eiffel Tower. It remains true that as a matter of logic, there is predictability in the sense that zf the same circumstances and antecedents appeared again, and if the same causal laws operate, then the same effects will occur. But methodologi- cally, there is no room for the pursuit of prediction in this sense, since we know that as a matter of fact the same laws will not operate in future and the same initial conditions will not recur. This, then, is one of the ways in which long-run predictions will run into trouble; and it must be noted that it 207 is not analogous to anything at all in the case of the natural sciences. For in the example we have chosen, the reason why the prediction itself alters the future circumstances depends upon people coming to understand why that prediction would have been successful; that is, the prediction alters subsequent behaviour, not in virtue of an independent status as a causal antecedent, but because people come to under- stand the prediction, and it then affords them reasons for behaving differently. This can only be the case where the phenomena in question have theories about their own actions and can sensibly be said to understand explanations of their own behaviour; and the only such phenomena of which we have any knowledge are human beings. The importance of this point will emerge a little later when we discuss self- validating and self-defeating predictions. Since it seems that predictions offered as accounts of what will happen no matter what are of no use as guides to action, or else are used as guides to action at the cost of their status as unconditional predictions, what of those conditional predictions which are used to tell us what would happen if no one intervened? It is plain that at an everyday level, we employ such predictions incessantly and essentially - though it should be noted that their scope is usually limited. (18) Thus we warn each other that the house will fall down about us unless we get the roof fixed, or that we will run down the little old lady on the crossing unless we slow down at once; and generally, economists, public health officials and a thousand and one inspectors and advisers tell families, governments and private persons what causal sequences are in train, precisely so that they will be able to intervene in them. More and more detailed knowledge of such causal sequences makes us more and more able to intervene successfully in them, and thus brings events increasingly under our control. But it makes a considerable hole in the ambition to predict changes far into the future, since the causal sequences laid out in such 'predictions' tend to be wide open to human frustration and modification. If what we learn is thus almost always what is happening, and what will happen if no one intervenes, we obviously learn rather little about what will happen, interventions and all. This conclusion, though a blow to a determinist picture of social life, is in no way destructive 208 of our account of the need in explanation for the production of such causal sequences; nothing in that account required that such sequences should be shown to hold far into the future, nor that they should be immune to future interven- tion by human beings. This intervention is no matter for logical disquiet. It is, in principle, possible to stick to one's determinist guns, and argue that long-range prediction is not in principle impossible, even though it may be very difficult to achieve, for all that is required is that we should be able to predict the interventions as well as the initial sequences, and to assess the effects of such interventions. Logically, this is obviously the correct response, even though it raises the problems we came across in Chapter 5 about giving a determinist analysis of choice. For if we are to make such predictions, they must be predictions of what will happen no matter what, only one of the events which will happen no matter what will be our own interventions. But this analysis of intervention is most disturbing, since it is so at odds with precisely what is usually thought to distinguish interventions, namely that they are a matter of choice. (19} In what sense can intervention be a matter of choice on this analysis? If what a prediction 'no matter what' achieves is to rule out intervention to frustrate the course of events predicted, we seem to plunge into very deep water if we rule out our intervention in our own inter- ventions. Such a move could only be made by a very committed fatalist who was prepared to abandon the whole concept of choice as a muddle. And such a fatalist would plainly not share the practical aims which I have suggested as the common motive behind an attempt to offer such predic- tions. It is thus apparent that both as a practical matter and an intellectually plausible goal, the attempt to rival astronomy with the production of a predictive social science is of dubious value. Of course, none of this rules out predictions at short range, and it should be said at once that almost all contemporary social science does restrict itself precisely to this kind of prediction - and as we shall later see, to the establishment of trends of recognisedly limited predictive scope. But in connection with such short-range predictions arises one of the interesting differences between the social and the natural 209 sciences. It is one of the features of social life that some persons are in a position of so to speak intellectual authority, in that their word on the state of affairs obtaining or about to obtain tends to be accepted as the truth; their say-so gives people a good reason for believing that what they say is true. Now, an effect of this which has been noticed by many writers, and christened 'the self-fulfilling prophecy' by Professor Merton, {20) is that the assertion by such authorities that something will happen may make it happen. The converse case of the self-defeating or suicidal prophecy amounts to the case discussed above, where the utterance of the prediction gives people a strong motive for making sure that it comes false. The kind of example to which the notion of the self-fulfilling prophecy appeals is the following: a man who is regarded as a reputable financial expert may say that a share will rise in price, and his word that this will happen is enough to set in train that rush of buyers which will actually bring about the rise in price which he has forecast. (21) Again, it is a generally recognised duty of members of a government to lie about the likelihood of devaluation, when there are rumours of devaluation in the air; the reason is that the statement of an official that devaluation was to take place would start such a stampede out of the currency in question as to bring about an inevitable devaluation. Now, it is important to notice that in these cases, the way in which the statement of the prediction would bring about the events predicted is very different from the case raised in Chapter 1, where a loud shout that an avalanche is coming is enough to actually bring on the avalanche. The difference is that in the case of human agents, the agents themselves have to both understand what the prediction means and to believe it before it can have any effect on their actions. It is only because they believe what is said that they can use it as grounds for the decisions which add up to the prediction's being fulfilled. Ttte prediction, in Popper's terms, gives the situation a certain logic to human agents. But in the case where I simply shout loudly enough to set off an avalanche, it is only the noise of my uttering the prediction that is at issue. What was predicted is neither here nor there; it only happens to be the case that the noise in question was the noise of my uttering a prediction; any other noise of equal 210 loudness would have been just as causally effective. Equally, it follows that the prediction would not have had any effect on the avalanche, had it been only written or had it been whispered; but in the case where human agents are involved, it is plain that any mode of diffusing the prediction would be equally effective as a causal agency, so long as the one condition is fulfilled that people understand what is said and believe it to be true. For in the case of prediction in human affairs, it is the meaning of the prediction that matters and not its status as a physical occurrence; (22) and it does not just happen that the prediction is causally effective, in the way it just happens that a loud noise will start an avalanche, rather it is only in the light of what the prediction means that it is logically possible for it to give someone a reason for behaving one way rather than another. As we saw in Chapter 6, what we are elaborating here is in large part a conceptual connection. We have said enough to indicate the logical problems associated with the idea that determinism reigns in social matters and that long-term prediction is the goal of social science. But before leaving the matter, I .should like to illustrate very briefly the practical consequence of these difficulties. The Marxist socialists of the German Social- Democratic party - the S.P.D. -before the First World War had grave intellectual, logical, ideological and policy troubles as the result of their ambivalence about the proper interpreta- tion of Marx's predictions of revolutionary historical change. In principle, the S.P.D. was committed to bringing about the revolutionary transformation of capitalist Germany into a socialist society; but the question that baffled many of the party's leaders was that of their role in this change. If Marx had been right in saying that the changes were inevitably on their way, it seemed to many of them that their role could not be to promote violent upheavals but rather to wait for the coming of socialism in the fullness of time, and to prepare themselves to be socialists in the new socialist order. Plainly, no one leading a mass working-class party like the S.P.D. was going to propose a policy of total inactivity as the only logical consequence of Marxian determinism; none- theless, a book like Professor Gay's 'The Dilemma of Democratic Socialism' (23) shows just how much the leaders 211 of the party felt themselves in a cleft stick. Were they to try to stir up a working class which, on all the available evidence, did not want a violent revolution, but was happy to make steady progress under the Bismarckian welfare state? Or were they to press through parliament and the organisations of the working class for the limited reforms of the status quo which the working class did want, in the hope that these would somehow cumulatively become the revolutionary transforma- tion predicted by Marx and Engels? Of course, the dilemma of policy is that faced by all socialist parties in industrialised countries, and there is no easy solution of it. But that is not the point. The point is rather that the sociological beliefs of the S.P.D. made the resolution of the dilemma impossible, since they seemed to rule out a policy which could fulfil the incompatible requirements of being revolutionary enough - i.e. activist enough - to be plausible as a policy, and yet fatalist enough - i.e. inactivist enough - to meet the logical requirements of a determinist Marxism. It can, of course, be said that this logical hiatus is not a sociologically adequate explanation of the S.P.D.'s failure, since plenty of people have been moved to action precisely by believing such logically incompatible things as that the victory of their cause was quite inevitable, and also that their fullest efforts were essential to its triumph. This is (up to a point) true, but it is a truth of less importance once a party or a group of any sort is committed to a policy which supposedly relies not on messianic faith but on a scientific assessment of the possibilities of social and political life. For in such a situa- tion, and this was the S.P.D. 's situation, intelligent, secular politicians will find themselves in a dilemma over policy. And no one who reflects on the contributory role of the S.P.D.'s weaknesses in allowing the collapse of the Weimar Republic and the subsequent accession of Nazism to power is likely to think its policy dilemmas unimportant. Thus far we have been sceptical about the goal of long- term prediction in social science. Yet of course there are examples of highly successful short-run predictions - predicting election results from pre-election polls, (24) predicting crime rates for short periods ahead, or birth rates and the like. A certain amount of economic forecasting and planning runs along similar lines, though, logically, it tends to 212 be a mixture of reported intentions, extrapolation of present trends, and statements of an exhortatory kind as well. But what we ought to notice about such short-run predictions is that in a great many cases they are not susceptible of much in the way of causal explanation, and rely not on causal laws- at least in our present state of understanding - but on the mere extrapolation of trends. Once we have made clear this distinction between trends and laws we can see whether there is any substance in Popper's claim that the addiction of many philosophers, historians and social scientists to the goal of large-scale prediction rests on confusing laws and trends. (25) The existence of unexplained but successful predictions is common knowledge. We have mentioned already the nineteenth-century discovery of the remarkably steady incidence of crime in France, made by two separate observers. The discovery was thought at first to be the beginnings of a truly scientific criminology; although, no doubt, individual criminals acted for all sorts of reasons, nonetheless, the results of these all but random events could be forecast with considerable accuracy. But the delusiveness of these hopes was soon realised, for it was seen that such uniformities in how many people committed crimes brought one no nearer discovering why people committed crimes, and thus no nearer knowing the causes of crime; all one had found out about was the rate of incidence of the effects. Such attempts as Lombroso's to explain criminal behaviour in terms of genetic defect and racial throwbacks were no doubt mistaken about the facts, but they were at any rate along the right logical lines in looking for the causal generalisations which would analyse the trends thrown up by Quetelet and Guerry as the outcome of the causal factors involved, operating in the circumstances of the nineteenth century. What criminologists need to know is what motivates criminal behaviour, what sort of upbringing makes a person vulnerable to temptation and so on. With this information to hand, they would he in a position to explain both the trends that existed at any particular time and changes in them, because they could explain both the causation of individual criminal actions and that of the gross statistical regularities resulting from these particular acts. (26) It is only by being able to understand the causal processes that underlie trends 213 that we are in a position to know how safely we can extra- polate a trend into the future. For a trend is not a statement of a lawlike kind at all; it is logically on a level with singular statements, and is indeed simply equivalent to a summary of singular statements: to say, for example, that the trend in petty theft is steadily upwards is only to say that over the period in question more crimes of petty theft were committed in period tn ... tn + 1 than in the period t n- 1 •.• tn. To plan a long way ahead on the basis of such a trend is plainly foolish in the absence of any understanding of the underlying causes. Economists and businessmen alike realise how very unwise it is to estimate future income on nothing more substantial than an unanalysed trend; and where they refer to trends in economic life, it is almost always with a suggestion that they wish to be committed for no great length of time ahead. In recent years the trends in population growth have offered some spectacular examples of the dangers of extrapolation, partly because the nature of birth rates is such that relatively small changes in the current birth rate will extrapolate to remarkable changes in future population over the long run, in just the same way as small changes in compound interest rates will make spectacular differences to one's savings over a long enough period. But it is notorious that birth rates change suddenly, and in the present state of our ignorance inexplicably. In fact when the ceremony to mark the birth of the 200 millionth American was held recently, it was estimated in various quarters that the true figure was some two million less; the trend on which the planners of the ceremony had relied, and on the basis of which the computer counting the babies to the 200 millionth American had been programmed, had reversed itself some time before the ceremony. (27) At a more domestic level, everyone who has been tempted to stake his earnings on a horse knows that there is a great difference between relying on a horse's winning streak - which is betting on a trend - and knowing about its form, breeding, condition and so on - which is betting on a set of causal sequences. If, for example, he could discover that the horse was light on its feet and thus particularly good in wet going, he would hesitate to risk his all if the weather became warm and dry. Similarly with birth rates, if we knew that the desire to have more children was a 214 response to increased well-being, rather than a response to absolute levels of well-being, we should not expect a high birth rate to persist either into a slump or into a long period of settled prosperity. In other words, we can rely on trends for predictive purposes only where we can explain how the trend is caused, and thus estimate its ·reliability. And this again explains why it is that we can safely rely on unexplained trends only over rather short periods - it is only over short periods that it is at all safe to assume that the causes of which we are ignorant have not altered in any significant way. But are we to believe that this not very problematic distinction between trends and laws was overlooked by those who are rightly regarded as the founders of the modern social sciences? There is no place here for a review of the enormous nineteenth-century output of work on the trends then visible in the fields of economic, political and social change; and such a survey would in any case be superfluous, since sociology today is quite sufficiently concerned with its origins in the responses of these theorists to the rapid social changes which took place before their own eyes. What, however, we can simply enough do here is explain what the argument hangs on. If it was possible to see in the work of Marx, Mill, Comte, Durkheim or Webet: an uncritical projec- tion into the far future of the changes visible in 1840, or 1860, or 1880, then there might be grounds for thinking that such confusion had taken place. For it might then be thought that they had failed to recognise the difference between saying that such and such changes have been taking place, and saying that whenever such and such a change occurs it has such and such consequences. For only this latter causal law would allow any kind of prediction - given that our earlier requirements as to the law's universality and the prediction of the initial conditions could be met. We may agree that these thinkers get into difficulties of one kind and another, but see these difficulties as the result of their doubts about what causes were really operating at the time they wrote, or how reliable the future operation of these causes was, and how profitable would be attempts to control their operation. For if they were well enough able to make the distinction between laws and trends, they were equally able 215 to make the distinction between modifiable and unmodifi- able causal sequences - and thus between the trends we can readily reverse and those which we cannot. It is obvious that in everyday life we use these distinctions quite unself- consciously. For example, we may with many contemporary sociologists see that our lives are becoming increasingly suburban, and that the distinctions formerly existing between urban and rural life are melting equally into suburbanism: this is to talk of a trend; it is a trend which can be partly explained in terms of our desire for fresh air, for living space and ease of movement for our children and so on; thus explained, we know that it is a trend which could in principle be reversed if we were able to forgo these desires, but we see that we are not likely to do so, and thus that the suburban trend is likely to go on. In this there is nothing very logically complex. And in essence, such a commonsense account is exactly what Mill for one offers in his essay on 'Civilisation' (28) - and the same defence can be made of the writings of Marx and Comte, though with allowances for the latter's curious beliefs about which causes were at work; while writers such as Weber and Durkheim were methodologically intensely cautious about prognosis. When Mill accounts for the _tendency of masses to predominate over individuals and for uniformity of manners to permeate society, both geographi- cally and through the class structure, he explicity claims that he is describing a trend. For in analysing the causes of this trend he inquires what we should have to do to reverse it or at any rate modify it. Since the causes seem to lie in the features of life which have also caused the industrial revolu- tion and its attendant changes, Mill offers little hope for a substantial reversal of the trend; but this is not because it is a universal causal law, which it makes no sense to try to alter, but simply because the costs of trying to alter a trend so bound up with the increasing prosperity of the age would be immensely high, and it would thus be foolish to try to reverse the trend. It is not that no matter what, the processes of industrialisation and the like must continue, but that their reversal would be intolerably expensive; but Mill insists as firmly as he can that while major reversals may be out of the question, it is quite possible to modify the side-effects of the 216 trend, and he encourages us to do this. Much the same analysis can be given of the seemingly more awkward case of Marx's insistence on the inevitability of the end of capitalism. (29) What Marx shows is simply that both intervention and non-intervention will equally bring to an end nineteenth-century capitalism of a pure and unregulated kind. In some ways, Marx's case is almost too much a matter of common sense. For often it boils down to saying that if capitalism is allowed to continue under no government surveillance, then increasingly violent booms and slumps will equally continue unchecked, for reasons which any classical economist would be forced to agree with. If this process goes on unchecked, then chaos and revolution will follow. But if there is to be intervention on a scale commensurate with the problem, this will require such a degree of government control and regulation as to amount in its turn to the revolu- tionary overthrow of capitalism from a different direction. Thus capitalism was caught in a cleft stick, and in that sense its end was inevitable - though its endings might be very different. It is of course quite likely that Marx was wrong about the processes at work, and that he oversimplified the range of alternatives open in the latter part of the nineteenth century; but this is to say that he relied on causal generalisa- tions that turned out eventually to be false; and this is a totally different charge from saying that he mistook the logical framework of a correct explanation. It would be no comfort to a confirmed Marxist to have Marx saved from charges of logical confusion only to have him condemned on charges of factual error; but this is immaterial here, for we are not concerned with the latter charges at all. All we are concerned to ask is whether it is those charges that are appropriate rather than the charges of logical error; and on the face of it, the logic of his account is impeccable. As we have said all along, if the goal of prediction fails, it does so because the facts of social life defeat it; this would be enough to account for any failures of which Marx may be accused; the charge of philosophical confusion here is both implausible and beside the point. The desire to predict the future might well have led Marx, as it has led many other writers, to overestimate the importance of the facts which he particularly had concentrated on, to the neglect of others 217 more influential than he believed - but this is to do a possible job less than perfectly, not to set out to do an impossible job. In view of the immense difficulties under which sociologists have always laboured, it ought to be plainly said that were they at all times totally committed to observing the distinction between the extrapolation of trends and the establishment of causal laws, they would inevitably make plenty of errors both in the one task and in the other.
NOTES
1. H. D. Lasswell and D. Lerner (eds}, 'The Policy
Sciences' (Stanford University Press, 1951). 2. Popper, 'Poverty', pp. 105-30. 3. Hempel, in 'Readings in Philosophical Analysis', p. 465. 4. Strawson, 'Logical Theory', pp. 82-90. 5. Nagel, 'Structure of Science', pp. 450-9. 6. Hempel, in 'Readings in Philosophical Analysis', p. 462. 7. N. Hanson, 'On the Symmetry between Explanation and Prediction', 'Phil. Rev.' (1959). 8. Brown, 'Explanation in Social Science', pp. 158-60. 9. D. E. Butler and D. Stokes, 'Political Change in Britain' (Macmillan, 1969). 10. Plato, 'Republic' p. 263. 11. Aristotle, 'Nicomachaean Ethics', (Oxford University Press) sec. 3. 12. St Augustine, 'City of God'. 13. R. C. Tucker, 'Philosophy and Myth in Karl Marx' (Cambridge University Press; 1961) pp. 21-7. 14. Saint-Simon, 'Selected Writings', ed. F. Markham (Blackwell, Oxford, 1952} p. 18. 15.J. S. Mill, 'AugusteComte and Positivism' (University of Michigan Press, 1961). 16. Mill, 'Logic', bk vi, ch. ix. 218