Contextualism, explanation and the social sciences

Abstract

Debates about explanation in the social sciences often proceed without any clear idea what an ‘account’ of explanation should do. In this paper I take a stance—what I will call contextualism—that denies there are purely formal and conceptual constraints on explanation and takes standards of explanation to be substantive empirical claims, paradigmatically claims about causation. I then use this standpoint to argue for position on issues in the philosophy of social science concerning reduction, idealized models, social mechanisms, functional explanations, inference to the best explanation and interpretive understanding.

Acknowledgements

Thanks to Don Ross and two anonymous referees for helpful comments on earlier drafts of this paper.

Notes

The best contemporary source is Michael Williams (1996). David Lewis (1996) and others also advocate what they call contextualism about knowledge. Their approach, however, is part of the project of finding necessary and sufficient conditions for conceptual analysis, something directly at odds with the kind of contextualism added here. Other sympathetic work includes Sklar (1999). Obviously early work identifying contexts in explanation (Garfinkel 1981; van Fraassen 1980) is also an inspiration, though its points were more narrow in scope.

The best argument of this sort is perhaps Stich (1990); see also Lakoff and Johnson (2001).

Here I am echoing Koertge's (1992) concern that an account of explanation ought to do work in clarifying controversies in philosophy of science and science, and that conceptual analysis for its own sake may not have such fruits.

A helpful elaboration here is that the answers to questions need not be entirely ‘in the head’ of any given individual—they may be distributed (see Wilson and Keil 1998). This is in line with the contextualist's anti-essentialism about theories.

Some (Sintonen 1999) make the questions inseparable by making explanation an ‘act’.

Another relevant factor may be how ‘deep’ we require the explanation to be. Some (Schurz 1999) argue that what is doing the explaining needs to be better explained than what it is explaining. Counter-examples to this as a necessary condition are easy to find (Weber and van Dyck 2002). But even if we give up the conceptual analysis paradigm, we can still grant that in some contexts this requirement is compelling.

One might then argue that unification tells us which causes explain, even if it relies on causal notions. But then as a universal criterion the problems discussed next emerge.

See Ross et al. (forthcoming) for a careful attempt to spell out the priority of physics.

The flagpole problem comes from the example of deriving the height of a flagpole from the length of its shadow and mathematical and physical laws which is an apparent counterexample to the N-D account of explanation, since the derivation does not explain the height of the flagpole.

This problem is really just an instance of the shadow and flagpole problem for N-D accounts of explanation, namely deriving does not suffice to explain.

Smith (2002) is a good guide here.

Henderson (1993) is a thorough argument along these lines.

A classic case of treating explanation this way is the use of R-squared statistical measure in the social sciences as a measure of explanatory power.