Saturday, November 3, 2007

Positivism and social science

There is a strong current of positivism in contemporary sociology --in fact, one might say this is the dominant paradigm. Other paradigms exist -- feminism, Marxism, comparative historical sociology, and ethnographic sociology, to name several. But the claim of science is generally couched in terms of a positivist theory of science and inquiry. This is unsurprising, in that several of the founders of sociology (Comte, Mill, and Durkheim in particular) were most emphatic in asserting the necessary connection between the two ideas, and Comte invented both "positivism" and "sociology" as modern terms.

The core assumptions of positivism include these: that social science is identical in its logic to natural science; that science involves the search for general laws about empirical phenomena; and that discovery and explanation depend upon a rigorous empirical scrutiny of the phenomena under question. Positivism is doubtful about the role of theory, preferring instead to make do with empirical observations, classes of empirical phenomena, and generalizations across classes of phenomena. Finally, positivism is dubious about the reality of causal connections between empirical phenomena.

It is true that science requires rigorous empirical inquiry. But much of the rest of the positivist program turns out to be badly suited to social science research and explanation. This is so for several reasons. First, social phenomena do not fall into fixed and distinct "types", in which the members of the type are homogeneous. We can generalize about "water", but not about "revolution", for the simple reason that all samples of pure water have the same structure and observable characteristics; but not so for all "revolutions". The category of "revolution" is not a "kind", and we should not imagine that we can arrive at a set of necessary and sufficient conditions for membership in this group.

Second, there are few lawlike generalizations about social entities and processes (if any at all!). Each revolution, for example, proceeds according to a historically specific set of causes and circumstances. And there are no genuinely interesting generalizations across the whole category.

Third, it is important in social science to formulate hypotheses about unobservable mechanisms of causal interconnectedness. So "theory" is an important component of social-science thinking and the sociological imagination.

Finally, explanation in the social sciences requires that we identify the causal mechanisms that connect one kind of social circumstance with another. If we believe that improved transportation causes a change in habitation patterns, then we need to be able to provide a hypothesis and analysis of what the social mechanisms are that create this result.

Positivism is a poor guide for social science inquiry. Instead, we need to approach social science research with a readiness to find contingency, heterogeneity, path-dependence, and particularity among the phenomena that we study--corresponding to the plasticity of human institutions and human agency.

1 comment:

Gustavo Lacerda said...

It's one thing to prefer empirically-grounded, but to be skeptical of theory for its own sake sounds silly (the more formal the theory, the sillier this skepticism). I'm curious how this philosophy has affected the practice of social science.

The reluctance to talk about causality also seems like a big mistake. It seems likely that the positivists are responsible for this dogma in orthodox statistics.

The limitations that you describe also apply to complex systems in the natural sciences.

Here's a vaguely related quote from Patrick Suppes's autobiography: << Knowledge of meteorology has stood me in good stead throughout the years in refuting arguments that attempt to draw some sharp distinction between the precision and perfection of the physical sciences and the vagueness and imprecision of the social sciences. Meteorology is in theory a part of physics, but in practice more like economies, especially in the handling of a vast flow of nonexperimental data.>>

<< The category of "revolution" is not a "kind", and we should not imagine that we can arrive at a set of necessary and sufficient conditions for membership in this group. >>

This sort of precise logical delineation is not necessary for a concept to be useful in science. Categories can be defined by clustering too.

My skepticism of claims involving the concept of "revolution" are about their ability to make testable predictions, of the sort objective enough for us to make bets on.