First, the favorable part of the story: positivism puts forward two ideas about conceptual clarity and empirical rigor that surely need to be a part of any intellectually sound effort to understand society, or to contribute to social science. Our concepts need to make sense (by some criterion of sense-making), and our assertions need to be supportable by some combination of empirical evidence and logical inference. These amount simply to the requirement that science should be rationally articulated and rationally justified. These are aspects of the epistemology of science advanced by the progenitors of positivism -- for example, Mill, Comte, the Vienna Circle, Schlick, Carnap, Hempel -- that I, for one, do accept. And if this were the full extent of positivism, then it would be hard to be anti-positivist.
But positivist social science makes several additional assumptions about social knowledge that are untenable, in my view.
First is naturalism -- the idea that the social and behavioral sciences should have the same structure and logical characteristics as the natural sciences. Chemistry and physics -- especially the classical versions of these sciences -- have a unified hypothetico-deductive structure; they discover laws of nature; and they derive the observable features of the domains of phenomena they encompass. Naturalism postulates, therefore, that sociology, economics, or psychology should have the same logical structure, because that is what "science" requires. John Stuart Mill clearly presupposed this assumption in his discussion of the "moral sciences."
Second, relatedly, is the unity of science -- the idea that ultimately all scientific theories should be subsumable under one "most fundamental" master theory. This assumption brings with it the idea of reductionism; higher-level sciences (psychology) should be reducible to lower-level sciences (neurophysiology). And "reducible" means "derivable from given suitable bridge definitions and laws". (This topic was central for the Vienna Circle logical positivists.)
Third is an assumption about methodology, to the effect that measurement and quantification are essential aspects of scientific knowledge. So quantitative statements and theories are preferable to qualitative or descriptive statements; and the goal of a social science should be to discover a set of variables within the domain of investigation that can be observed, measured, and counted. This is a different aspect of the unity-of-science doctrine: the idea that there should be one privileged method of discovery and presentation for the social sciences. Where does this assumption come from? In part, it seems to derive from the physics-envy associated with naturalism; but perhaps there is also a Platonic dimension as well -- a preference for mathematics over descriptive or interpretive language.
Fourth is an assumption about explanation, regularities, and laws. The assumption here is that explanation requires the discovery of law-like generalizations about the domain of phenomena encompassed by the scientific field. This assumption has two components: the idea that a well-defined domain of investigation must somehow embody a set of regularities, perhaps disguised by the noise; and second, that explanations within the domain of individual events or patterns of events must take the form of a derivation of the explanandum from the general laws mentioned in the explanans. Carl Hempel and J.S. Mill agree about this premise.
Fifth is an assumption about causation -- that causation is a feature of statistical relationships among variables rather than a feature involving causal necessity or causal mechanisms. This is a Humean approach to causation, and it leads positivist social scientists to restrict their attention to causal regularities rather than looking for real causal mechanisms.
Finally, there is a sixth premise that has also created debate but seems less intrusive to the practice of innovative social science -- the insistence on the fact-value distinction. "Positive" science has to do with the discovery of facts, whereas ethics or policy stidies have to do with values.
Do these assumptions necessarily travel together? Not necessarily, though there are some internal logical connections among them that make it more difficult to imagine them standing completely independently. But it appears to be a characteristic of the observed sociology of science for an important stream of twentieth-century social science research, that these features are clustered together. And many critics argue that these assumptions have created blinders for social-science researchers, limiting their originality in theories, concepts, and explanations of the social world.
Critics of positivist social science ask us to consider a broader space of possibilities for research and theory formation in the social sciences. Taking the premise of scientific rationality as a given, what would a philosophy of social science look like that questioned the other premises on this list? What is a "post-positivist realism" for the social sciences?
- It is realist about causation; it affirms the scientific validity of seeking for real social mechanisms.
- It advocates for a conception of scientific explanation that hinges on the discovery of real causal connections among features of the social world.
- It is pluralistic about method; it acknowledges that there are multiple rationally supportable methods of inquiry in the social sciences, and multiple forms that social-science knowledge can take.
- It is even-handed among quantitative, qualitative, comparative, and narrative approaches to social inquiry and social explanation.
- It is anti-reductionist and anti-naturalistic: it does not presuppose that various areas of the social sciences should be reducible to some other, more fundamental scientific theory; and it does not presuppose that the social sciences should resemble the natural sciences.
- And, finally, it is fully committed to the positive features of rationality that were mentioned above: the scientific virtues of conceptual clarity and empirical-rational justification for scientific beliefs.
This set of alternatives opens up the space of the social sciences quite dramatically; it permits a wide and pluralistic range of inquiries to proceed, without the requirement of theoretical or methodological unity. And this frees researchers to arrive at accounts of their domains of research that are well suited to the particulars of these domains.
In a later posting I will come back to an important contribution to this debate, George Steinmetz's The Politics of Method in the Human Sciences: Positivism and Its Epistemological Others.