Saturday 11 June 2011

Statistics in Social Science and Causal Reality

The issue over the nature of causes is all about the defensibility of explanations. Statistics in social science help to establish regularity in observations. By doing this, it is hoped that social science can be seen to be compatible with successionist causal theory (Humean theory) that was so effective in physical science. Hume's assertion that causes were constructed in the light of observation of regular successions of events seems reasonable in physical science, but in the open system of social science, it was always problematic. So statistics comes to the rescue!

But this is problematic for a number of reasons. Firstly, following Kuhn, the closed system of physical science wasn't that closed: the scientific community, ideas of other scientists, instruments, etc all played a role in the construction of causes, so that the regularity that was observed was tensed and situated in a social situation which was not accounted for in the construction of a theory. In short, 'my head' was not just 'my head' - it was mine and the heads of others in history, the ideas of the time, the ideas that influenced the design of instruments, and so on. What we judged we judged together, but assumed it to be individual judgement that was reproduced in the light of phenomena, and therefore those judgements were defensible on the basis that they were agreed. But at the deep ecological, historical, political and psychological level, their agreement was just another level of 'something going on'.

Then there's the issue of what scientific explanations seem to do. Whilst they were concocted in closed system experiments, they had explanatory power in open systems too: things work in the world as well as in the laboratory. What does that tell us? Possibly that whatever was imputed to exist through experiment and construction of causes actually did exist in reality independently of any construction of causes. Therefore something must be happening which is outside peoples' heads and which somehow we have managed to grasp hold of.

In social science, the same is true: things happen in the world, and deep down, artists and poets know what it is. Social scientists are less confident to express what it is, and so they turn to statistics. What does statistics do? Statistics supposes that for a science of society to be possible, regularities of observation must be possible, so it finds a way of producing regularities in an open system.

The central question is "does the possibility of a social science depend on regularities?". I think the answer is no. The possibility of social science depends on explanatory power, and that depends on understanding the nature of explanation and understanding. In physical science, mechanisms are likely to exist outside the close-system experiment. Similarly, in social science, mechanisms are likely to exist beyond any local regularity or statistic. The challenge is "how can we come to rational and defensible knowledge about it?"

Each thing that happens tells us something about what's going on. This is probably what scientists really do - they are like detectives, piecing together clues and constructing possible ways in which those clues might be connected. The thing that they're describing is however a real thing, and they're making guesses as to how it might work. Social scientists are the same, but the mechanisms are more complicated, and they are always part of them (in fact so are the physical scientists!). But with enough evidence, and enough creativity in postulating what might be going, a science of society is no less possible than physics.

No comments: