Realism and empirical reasoning

How should a realist think about confirmation of hypotheses and theories about real social structures and properties?

Here is what the Ur-realist Roy Bhaskar has to say about the object of scientific knowledge in his approach in A Realist Theory of Science:

[CR] regards the objects of knowledge as the structures and mechanisms that generate phenomena; and the knowledge as produced in the social activity of science. These objects are neither phenomena (empiricism) nor human constructs imposed upon the phenomena (idealism), but real structures which endure and operate independently of our knowledge, our experience and the conditions which allow us access to them. (14)

So how would a realist confirm or justify the claims he or she want to make about these real underlying structures and mechanisms? What kinds of evidence and modes of reasoning are available to allow us to conclude that a given statement is true (or false)?

This is an important question, and it poses difficulties for critical realists. The problem arises from the rhetoric of CR and its staunch anti-positivism. The most obvious answer to the question takes us to the logic of observation, experiment, and testing through derivation of indirect implications of hypotheses. We are led to think immediately of Mill's methods and the logic of causal inference as primary modes of scientific reasoning. And this begins to sound a lot like -- the hypothetico-deductive model of confirmation, a key premise of neo-positivism (link). But critical realists are a little allergic to this answer.

So how can we sort out the reasonable and uncontroversial core of our ordinary intuitions about scientific reasoning from the full baggage of logical empiricism? One crucial difference is the neo-positivist presumption that scientific knowledge must take the form of general statements of lawlike regularities. CR focuses instead on singular statements about causal powers and mechanisms. And testing of singular causal statements has important differences from testing comprehensive unified theories. If we believe that vitamin C has the causal capacity to inhibit infection by the rhinovirus, it is fairly straightforward to design an experiment or observational study that helps to evaluate this belief.

Another important difference is the centrality of the observation-theoretic distinction for the neopositivist theory of science. Positivist and neopositivist philosophers of science often accepted the Duhem thesis about the nature of scientific knowledge (The Aim and Structure of Physical Theory) -- the idea that scientific knowledge takes the form of a unified body of theory that can only be evaluated as a whole, and that this system unavoidably includes the use of theoretical concepts that are inherently unobservable. The CR approach, on the other hand, rejects a strict distinction between social reality and social observation. There is nothing in social reality that is in principle unobservable.

Bhaskar places the idea of a scientific experiment at the core of his justification of the theory of transcendental realism. But surprisingly, he does not discuss the epistemic importance of experimentation. What does a successful experiment contribute to knowledge? The intuitive answer is fairly obvious: an experiment allows us to assign particular causal effects to particular interventions. C in the absence of I leads to O; C in the presence of I leads to P. Therefore I causes P. The successful experiment gives us empirical and logical grounds for believing that "I causes P". But perhaps surprisingly, Bhaskar does not pay any attention to the epistemic value of experimentation, or the role this method plays in inferences about real causes of changes in our world. Bhaskar prefers to argue that experimentation contributes to the ontology rather than the epistemology of science.

So, once again -- how should a critical realist offer an empirical justification for a claim like this: structure S has causal powers P and tends to bring about outcomes O?

It seems to me that the answer is not esoteric. There is no better way of establishing the credibility of a social hypothesis than we can find through the familiar modes of causal inquiry, detailed sociological investigation, and an accumulation of knowledge about concrete aspects of the social world. A realist researcher is advised to pursue familiar approaches to the problem of assessing the truth or falsity of hypotheses about causal properties of social entities. Do we believe that a certain kind of industrial organization has the tendency to produce a higher-than-average level of industrial accidents? Then collect observational data about how this organizational form functions; how it is internally regulated; and how its internal composition gives rise to a propensity for industrial accidents. This is the kind of research that Charles Perrow carries out in books such as Normal Accidents: Living with High-Risk Technologies and The Next Catastrophe: Reducing Our Vulnerabilities to Natural, Industrial, and Terrorist Disasters. Hypotheses about social structures and their powers can be investigated piecemeal, and we can tease out the causal properties that they possess through familiar methods along the lines of Mill's methods (link, link).

Take two ideas about social structures in particular: the idea of an economic structure and the idea of a business corporation. Both these ideas can be articulated in reasonable detail and then investigated through piecemeal empirical inquiry. Consider the economic structure: what are the property relations through which labor and other resources are managed and controlled in specific instances of capitalist economies? How do these relations work in concrete detail? What variations exist? What failures do these relations experience? This is the kind of materialist historical research that is conducted by Charles Sabel and his colleagues offer in World of Possibilities: Flexibility and Mass Production in Western Industrialization.

Or take a specific corporation -- the IBM corporation, for example. Here again, it is unproblematic to identify many independent avenues of research through which the relations, properties, norms, and patterns of behavior that make up the corporation. There is nothing occult about the corporation. Do we think that the power of management distorts the decision making of the corporation by favoring executives over shareholders and employees? It is straightforward to think of numerous independent ways of empirically investigating this hypothesis. Do we believe there are significant principal-agent problems among employees and corresponding solutions in the relations and roles embodied in the organization? Likewise these hypotheses can be empirically investigated.

So it seems most reasonable to argue that empirical reasoning looks the same for critical realists as it does for any other empirical scientist. There is no distinctive critical-realist model of empirical reasoning.