Lecture One: The Aspiration for a Natural Science of the Social
Explanation
These lectures presuppose that the primary task of science is to explain. This does not mean that the only task of science is to explain; description and classification are acknowledged to be important aspects of science, and not everyone thinks they are important only as tools for explanation. More importantly, not everyone thinks that social science in particular is geared towards explaining the phenomena it studies; some writers appear at least to think that interpreting is not explaining. I shall argue that the appearances are deceptive and that even they are committed to the view that explanation is the name of the game; I shall also try to show that it is (usually) possible to accept the substantive claims of such writers while (often) rejecting the philosophical glosses that they put on them.This involves working our way through ideas about makes explanations explanatory, then through a number of claims about the form that explanations in the social science ideally take: functionalist, causal, interpretative, ratonal choice and so on.
What is the philosophy of science about?
This is itself a philosophical question, and answers range from ‘nothing’ to ‘policing the rationality of the scientific enterprise.’ The argument for ‘nothing’ rests on the perception that no physicist would be deterred by a philosopher telling him he was doing physics badly; conversely, physicists don’t care whether philosophers endorse their work or not. Standards, on this view, are internal to a discipline, or more broadly to a practice – so that I cannot, but a trained priest could, tell a new officiant how to celebrate the Eucharist. But, even a scientist who thinks that philosophers have nothing to contribute will think in addition that what she is doing is science and that doing science is interestingly different from all sorts of other activities. This presupposes at least that there is some form of science/non-science demarcation that is taken for granted in practice, but which can be elaborated on and justified. (Otherwise, there is the danger of extreme relativism, where anyone can say what they like about anything, and there is no consensus on what is true/false or unassessable.) The consensus view is: i) philosophers cannot tell scientists how to do their job – either natural scientists or social scientists – any more than they can tell car mechanics or street cleaners how to do their jobs; ii) they can elaborate an account of the scientific enterprise in much the same way that they can give an account of art, poetry, religion etc; iii) they can in a loose sense of the term ‘logic’ give an account of the logic of different sorts of explanation and understanding. The issues discussed this term fall into that broad category. One respect in which philosophy of social science is peculiar is ‘physics envy’ – or perhaps engineering envy.That is, the practical aims of social science have (largely) remained unrealised, whereas non-social technology has since the 18th century advanced at an accelerating pace. Understanding why the very idea of social technology is a fraught topic with no ready analogue in philosophy of physical science.
The task of science as the production of explanations;
To pre-empt some later arguments: to say that science sets out to produce explanations is too quick; sometimes, we are grateful for predictions – especially when they are warnings – even when they do not come with explanations attached. Reliable ways of predicting earthquakes, hurricanes, epidemics would all be very worth having even if they left us as puzzled as ever about what causes them. (Think of populations reacting to plague by escaping contact with the affected, or more obscurely, people fleeing earthquakes because birds fall silent.) Opinion polling is useful to politicians, but we are not altogether sure what makes people vote one way rather than another. And to pre-empt another famous argument, the non-isomorphism of explanation and prediction goes both ways: explanations do not always produce predictions; we may be good at explaining post mortem why someone has died, but still be unable to predict who will and won’t die. In any non-deterministic system, prediction over more than a short period will be impossible; and even a deterministic system is indeterministic if it is open to other, non-deterministic systems.
Different accounts of explanation, psychological and logical;
The question is what is an explanation; or what makes an explanation explanatory. The basic distinction is between accounts that make ‘explanatoriness’ a psychological matter and those that make it a matter of logic; ‘logic’ is not a wholly satisfactory way of drawing the contrast, but given the salience of the so-called ‘hypothetico-deductive’ account of explanation, it is more satisfactory than most alternatives. Psychological theories mostly focus on the effect on the person who asks for an explanation; and they range from the achievement of what is known as the ‘aha effect’ – familiar from Conan Doyle stories – to what one might call the ‘puzzlement reduction’ effect. A person who asks for an explanation does not see how (or why) something happens as it does; the explanation reduces that puzzlement; and may achieve the ‘aha, now I see’ response. There are some obvious objections to this theory: people may have their puzzlement reduced by all sorts of things other than an explanation, familiarity in particular; more importantly, the well-conducted mind has its (rational) puzzlement reduced only by a ‘real’ explanation. The psychological impact must be parasitic on the purported logical weight of the explanation. This is worth dividing into two aspects, the formal and substantive; the formal is a matter of whether if things are as the explanation says, the explanation would be sound; the substantive a matter of whether things are indeed as the explanation says. (For instance, anti-biotics combat bacterial infections only; if I give you anti-biotics and your ‘flu gets better, it is not because of the anti-biotics because ‘flu is caused by a virus. Formally, an explanation that runs ‘everyone treated with anti-biotics gets over ‘flu; Jones was treated with anti-biotics; so Jones got over ‘flu’ is fine; substantively it’s not. And this is true even though there are many true propositions about infections that are treatable with anti-biotics.)
Causal explanation and its analysis;
I will run through – this week briefly and next week less briefly – the main claims of the so-called ‘hypothetico-deductive’ account of explanation. It was for many years canonical; its pedigree goes back to Mill, to Whewell, and perhaps to Hume. One of its virtues is that it is built firmly on ordinary deductive logic; another is that it separates formal and substantive adequacy; a third is that it illuminates – if it doesn’t wholly explain – the nature of causal explanation; a fourth is that it shows the role of generalizations and especially causal laws in explanation; and it can – in Popper it is the main feature of the theory – show the importance of disconfirming experiments. The shape is: explanans explains explanandum by providing a valid deductive argument in which the explanans sentences entail the explanandum sentence(s); the explanans must contain at least one general statement; and the explanans must be true or at least well-confirmed. Now we need a few examples....
Causal laws and accidental generalizations.
Not all requests for explanation are requests for causal explanation: eg, a demand for a translation or an interpretation of a poem or play or novel. Requests for the explanation of events are almost always causal (some are requests for meanings); imagine Jones dead on the floor, or the broken window. We want to know: what made that happen? Ordinarily, we adduce a singular causal claim – the brick hit it, he swallowed cyanide; the usual h-d account is that we offer sketches of a full explanation. There are other views, eg, that we think of things as exerting natural powers.The standard empiricist view is that talk of natural powers is unacceptably metaphysical; causal connections are simply ‘whenever A then B.’ This requires an account of the difference between causal and accidental generalizations,which in most modern accounts has invoked the role of causal claims in supporting counter-factuals.The canonical claim is that the basic causal generalizations are laws;these strictly, are timeless and hypothetical, and have three major virtues: precision, universality, generality. That raises the nasty question: are there any laws in social science? Mill said not, for interesting reasons.
Is all this misguided?
The first version of naturalism clear enough to attack was Mill’s in System of Logic. It duly provided the target for – in particular – Dilthey, who argued that there was a categorical or conceptual distinction between the sciences of nature and the sciences of culture; the former rested on causal explanation, the latter on interpretative understanding; absurdly, Mill’s distinction between the physical and moral sciences – those where mental phenomena were not, and were, involved – which was supposed to show that both were natural sciences was translated as the distinction between Naturewissenschaften and Geisteswissenschaften, which was supposed to indicate a sharp break between them.
1