There is no such thing as a social science

Phil Hutchinson, Rupert Read, and Wes Sharrock (HRS) offer a nice defence of Peter Winch in their provocatively titled book. A lot of the book is aimed at critics of Winch, arguing that he is not a relativist, not an idealist, not conservative, and so on. There's also a chapter on the strengths and weaknesses of ethnography and ethnomethodology. I'm more interested in their reconstruction of Winch's argument.

On page one they make the point that they do not oppose either analytical rigour or "a programme of social inquiry which accords a deal of importance to the revision of its claims in light of further (relevant) observations". What they oppose is two things: methodological reductionism and substantive reductionism. The former is the claim that (p. 2):

There is an identifiable scientific method and this ought to be employed if one intends to make a claim to do something scientific.

The latter is the claim that (p. 2):

Social scientific findings are reducible to the findings of the natural sciences.

This second claim strikes me as obviously false. If there are laws about the relation between inflation and unemployment, for instance, then these are not going to be about atoms or chemicals in the brain. Indeed if there is to be a social science then such reduction surely has to be impossible, otherwise the alleged social science will really be just one or more of the natural sciences after all. The idea of a social science is that there is something distinct at the social level to be studied. Or so I would have thought.

Methodological reductionism is much more plausible.

The Oxford English Dictionary defines the scientific method as: "a method or procedure that has characterized natural science since the 17th century, consisting in systematic observation, measurement, and experiment, and the formulation, testing, and modification of hypotheses." 

Why couldn't psychologists and economists, for instance, observe, measure, experiment, and hypothesize in this way? HRS grant that social scientists can observe and revise their claims in the light of observations. If they can observe they can presumably measure. (Perhaps it's dangerous to presume too much, but I don't think anyone would deny that psychologists could, say, count the number of people who cheat or lie or help someone in such-and-such circumstances, and that is surely a kind of measurement.)

One problem is with the idea of experimenting in social studies. There are many problems with this. There is, for instance, the ethical aspect. We can't just play around with an economy or culture the way we might with inanimate objects or mice in labs to see what happens. Another issue is the problem of ecological validity, discussed by McManus. How can we be sure that what happens in a controlled, laboratory environment also applies in the real world? And if we don't control the environment but just study the real world itself, how can we be sure that the results we observe are caused by the factors we hypothesize as the cause and not others? The obvious thing to do would be to repeat the 'experiment' as many times as possible, but in the real world we cannot make this happen and there are always going to be potentially complicating factors. So we can hypothesize, observe, measure, and modify our hypotheses, but it is doubtful that we can really experiment in a useful way, i.e. a way that will tell us anything we really want to know.

Another problem is that the OED model of the scientific method is not very specific. What counts as systematic? What should be measured? How? What kinds of hypotheses should we be formulating? We can't just take what physicists do and then "do the same" with human beings. Or rather we could, but if we really do the same then we will just be doing physics with a foolish and probably unethical choice of equipment. If we decide to do the same thing but different, then the same how? and different how? These questions require decisions about what we want to do and how we choose to do it. The answers are not given by the scientific method or by the phenomena to be studied. We don't yet know, have not yet decided, what these are. It is not the case that there is an identifiable scientific method that can simply, without doing anything else significant, be employed or applied.

Two other points that HRS emphasize are the question of the identity of action and the nature of understanding. If we want to find laws of human behavior then we need to identify cases of the same thing being done. As Anscombe has pointed out, a man who is moving his arm up and down while holding a pump might be described as doing several different things: playing a tune with some squeaky machinery, exercising his arm, pumping water, poisoning people (if the water supply is poisoned), and so on. If we want to understand him then we need to know what he takes himself to be doing. He isn't murdering anybody if he doesn't know there is poison in the water. We could, I suppose, study his movements just as physical movements, but a) it will make a difference whether we study the movements of, say, atoms, muscles, or the air around his body (which science we are engaged in depends on this kind of thing), and b) we won't be doing social science in that case. If we care about his behavior as action then we need to know his intention. This is a complicated matter, as Anscombe's work shows, but it has to do with the man's conception of his action (what he takes himself to be doing, which depends on the concepts he has) and his reasons for acting in this way (why he is doing--or at least takes himself to be doing--what he is doing). We can think of reasons as a kind of causes, but they are very different from paradigmatic causes. Reasons can be good or bad, for instance. The same wind that causes me to lose my footing can similarly cause leaves to fly around, but nothing like my reason for taking a walk will ever move the leaves. Reasons only ever motivate rational beings (this is not a matter of chance) and can be evaluated. They belong to a different domain, as Paul Johnston puts it, from that of causes. If scientific explanation is causal (in the billiard ball sense of causal) then human action, qua human action, cannot be understood scientifically.

This brings us to the last point, the nature of understanding. To say that human action cannot be understood scientifically is not to say that it cannot be understood. We understand it (and fail to do so) all the time. It is usually obvious what someone is doing. If it isn't, we can ask them. And they usually tell the truth intelligibly. If we still don't understand then we might need to look around more. What is the context in which this behavior is taking place? Is it part of a game, or a religious ritual, or a play, or a way to frighten animals away, or what? It helps, as HRS point out, if we can relate the behavior to something similar that we do. If we can't then it might remain somewhat mysterious: Well, it's music of some kind, but what they see in it I'll never understand. There is no guarantee that we will always be able to make full sense of other people's behavior, but there's no guarantee either that they are bound to be and remain closed books to us. We have to look and see. That's something that of course we can do, but it isn't science in any significant sense of the word.