Friday, March 30, 2012

Hempel after 70 years

Carl Hempel published his sole contribution to the philosophy of history in 1942, almost exactly 70 years ago. The article is "The Function of General Laws in History" (link), and it set the stage for several fruitless decades of debate within analytic philosophy about the nature of historical explanation. Hempel argued that all scientific explanation has the same logical structure: a deductive (or probabilistic) derivation of the explanandum from one or more general laws and one or more statements of fact. Explanation, in Hempel's view, simply is "derivation of the explanandum from general laws." Here is the opening paragraph of the essay.
It is a rather widely held opinion that history, in contra-distinction to the so-called physical sciences, is concerned with the description of particular events of the past rather than with the search for general laws which might govern those events. As a characterization of the type of problem in which some historians are mainly interested, this view probably can not be denied; as a statement of the theoretical function of general laws in scientific historical research, it is certainly unacceptable. The following considerations are an attempt to substantiate this point by showing in some detail that general laws have quite analogous functions in history and in the natural sciences, that they form an indispensable instrument of historical research, and that they even constitute the common basis of various procedures which are often considered as characteristic of the social in contradistinction to the natural sciences. (35)
And here is the logical structure of such a "covering law" explanation, according to Hempel:
(1) a set of statements asserting the occurrence of certain events C1, . . . C, at certain times and places,
(2) a set of universal hypotheses, such that
          (a) the statements of both groups are reasonably well confirmed by empirical evidence,
          (b) from the two groups of statements the sentence asserting the occurrence of event E can be logically deduced. (36)
He is emphatic, moreover, in insisting that valid explanations in history must have this form:
We have tried to show that in history no less than in any other branch of empirical inquiry, scientific explanation can be achieved only by means of suitable general hypotheses, or by theories, which are bodies of systematically related hypotheses. (44)
Hempel concedes the point that few existing historical explanations actually look like this, with explicit law statements embedded in a deductive argument; but he argues that this shows only that existing explanations are elliptical, incomplete, or invalid. And often, he finds, what is offered as a historical explanation is in fact no more than an "explanation sketch" (42), with placeholders for the general laws.

What kinds of general laws does Hempel think that historians have in the back of their minds when they offer elliptical explanations? He refers to regularities of individual or social psychology (40), regularities of collective behavior ("groups migrate to regions which offer better living conditions"), or at the macro level, regularities linking growing discontent to the outbreak of revolution (41). Further:
Many of the universal hypotheses underlying historical explanation, for instance, would commonly be classified as psychological, economical, sociological, and partly perhaps as historical laws; in addition, historical research has frequently to resort to general laws established in physics, chemistry, and biology. (47)
This set of assumptions leads to big trouble for historical explanation if we accept Hempel's account, however, because it is hard to think of a real historical research question where there might be a set of social or individual regularities sufficient to deductively entail the outcome. Bluntly, the social and behavioral sciences have never produced theories of individual or collective behavior that issue in statements of general laws that could be the foundation for a covering law explanation. And given that social phenomena are formed by actors with a range of features of agency and decision-making, we have very good reason to think that this lack of regularities is inherent in the social world. The social world is simply not governed by a set of social or individual laws. Let's look at that point at several levels.

Individuals. The social sciences provide a good basis for advancing theories of agency, which in turn support certain generalizations about action. For example: People act out of self interest. People act morally. People pay attention to the example of others. People care about their families and friends. People follow charismatic leaders. People follow the precepts of their religious beliefs. People are emotional and short-sighted. People make decisions based on specific heuristics and rules-of-thumb. Each of these statements takes the form of a generalization. And each is true -- of some delimited groups of agents some of the time. But there is no generalization about agency that is true of all agents all the time. Rational choice theory attempts to provide a single theory of agency and decision making that replaces all of these variant grounds of action. But rational choice theory has proven notoriously unsuccessful as a foundation for explanation of a large and complex event -- war, revolution, economic crisis.

Groups. Here too we can identify some partial regularities: Groups tend to coalesce in action when they have prominent shared characteristics.  Groups are more prone to panic than individuals. Groups tend to fail to accomplish collective purposes. Groups are hyper-sensitive to racial and ethnic markers.  And so forth. It is evident that these are partial, tendential, exception-laden, and inexact; not at all like the generalizations that characterize metals, liquids, or proteins.

Organizations and institutions. What about mid-level social arrangements like labor unions, congregations, and terrorist cells? It's not that there aren't any generalizations to be had concerning items at this level; it is that there are too many, and they are highly contingent, conditioned, and contradictory. Certain types of organization are more prone to accidents than others.  This is true; but we have more confidence in our analysis of the most important features of the high-safety organization than we have in the corresponding generalization.  So there isn't a stockpile of laws that might be produced to apply to a social situation and then turn the crank and derive the deductive consequences.

Finally, what about large-scale events and structures -- wars, revolutions, civil conflict? Here too there are some generalizations that social scientists have asserted. For example: Democracies don't go to war with each other. War is made more likely when two powers have conflicts of interest over important resources. Wars create propaganda.  Revolutions don't happen when the general population is satisfied. But generalizations about these sorts of social entities too are bounded and unreliable. They are conditional, we recognize immediately that they have exceptions, and they don't permit prediction.

So the strong, governing generalizations that would be needed for a covering law explanation do not exist. As I argued a number of years ago, social regularities are phenomenal, not governing (link); they reflect characteristics of the actors rather than governing the behavior of the ensembles.  Does this mean that historical explanation is impossible?  No.  But we need to turn our attention from regularities to causal mechanisms and powers in order to see what a good historical explanation looks like.  A good historical explanation identifies a number of independent mechanisms and processes that are at work in a particular circumstance, and then demonstrates how these mechanisms, and the actions of the actors involved, lead to the outcome.

Doug McAdam, Sidney Tarrow, and Chuck Tilly advanced a boldly different approach to analyzing and explaining complex historical phenomena, with special application to social contention.  They rejected the idea that there might be "laws" of revolution, civil unrest, or ethnic cleansing. They argued instead that there are a number of recurring "social mechanisms" of contention that can be identified in many instances of contention, and whose influences can be traced out to result in the observed outcomes.  Here is how McAdam, Tarrow and Tilly proceed in Dynamics of Contention.

We begin with a question: What led normally accepting accepting African-Americans both in Montgomery and throughout the South to risk their livelihoods and their lives in support of civil rights? Recall from Chapter I that in the "classical social movement agenda" the following factors come into play: 
  • Social change processes initiate a process of change and trigger changes in the political, cultural, and economic environments. 
  • Political opportunities and constraints confront a given challenger. Though challengers habitually face resource deficits and are excluded from routine decision making, the political environment at any time is not immutable; the political opportunities for a challenger to engage in successful collective action vary over time. These variations shape the ebb and flow of a movement's activity. 
  • Forms of organization (informal as well as formal) offer insurgents sites for initial mobilization at the time opportunities present themselves and condition their capacity to exploit their new resources. Despite some evidence to the contrary (Piven and Cloward 1977), a large body of evidence finds organizational strength correlated with challengers' ability to gain access and win concessions (Gamson 1990). 
  • Framing, a collective process of interpretation, attribution, and social construction, mediates between opportunity and action. At a minimum, people must both feel aggrieved at some aspect of their lives and optimistic that acting collectively can redress the problem (Snow, et al. 1986; Snow and Benford 1988). Movements frame specific grievances within general collective action frames which dignify claims, connect them to others, and help to produce a collective identity among claimants. 
  • Repertoires of contention offer the means by which people engage in contentious collective action. These forms are not neutral, continuous, or universally accessible; they constitute a resource that actors can use on behalf of their claims (Traugott, et al. 1995). The use of transgressive forms offers the advantages of surprise, uncertainty, and novelty, but contained forms of contention have the advantage of being accepted, familiar, and relatively easy to employ by claimants without special resources or willingness to incur costs and take great risks.
That classical agenda made three enduring contributions to the study of social movements. First, it made strong claims regarding the close connection between routine and contentious politics, helping to reframe the study of social movements as the proper province of both sociology and political science. Second, calling attention to the role of "mobilizing structures," it represented a powerful challenge to the stress on social disorganization and breakdown in the older collective behavior paradigm. Third, it produced a credible picture of mobilization into social movements that was supported by a good deal of empirical evidence correlating the factors outlined above with increases in mobilization.

There are low-level generalizations offered throughout this series of statements. But all those generalizations are soft and exception-laden.  What MTT are interested in doing when they attempt to explain what they call "episodes of contention" is rather to identify the occurrence and interaction of a number of common mechanisms of contention.  And in fact, they explicitly repudiate the covering law model:
Our emphasis on recurring mechanisms and processes does not mean that we intend to pour all forms of contention into the same great mold, subjecting them to universal laws of contention and flattening them into a single two-dimensional caricature. On the contrary, we examine partial parallels in order to find widely operating explanatory mechanisms that combine differently and therefore produce different outcomes in one setting or another. To discover that third parties influence both strikes and ethnic mobilization by no means amounts to showing that the origins, trajectories, and outcomes of strikes and ethnic mobilization are the same, any more than identifying similarities in memory processes of mice and men proves mice and men to be identical in all regards. To discover mechanisms of competition and radicalization in both the French Revolution and in the South African freedom movement is not to say that the Jacobins and the African National Congress are the same. We pursue partial parallels in search of mechanisms that drive contention in different directions. Only then, and in Part III, do we examine how mechanisms combine in robust political processes.
Seventy years after Hempel's classic article, the covering law theory is now generally regarded as a fundamentally wrong-headed way of thinking about historical (and social) explanation.  Logical positivism is not a convenient lens through which to examine the social and historical sciences.  There is too much contingency in the social world. Rather than being the result of law-governed processes, social outcomes proceed from the contingent and historically variable features of the actors who make them.  So the attention of many people interested in specifying the nature of historical and social explanation has focused on social mechanisms constituted and driven by common features of agency.

(Renate Mayntz's discussion of causal mechanisms represents one of the best current treatments of the subject; link.)