Abstract

A set of techniques for identifying and estimating causal dependencies within a system--variously called Structural Equation Modeling (SEM), Structural Causal Modeling (SCM), causal inference, or just causal modeling--have been developed in the context of the statistical sciences, starting early in the 20th century but mainly over the last several decades, and are currently seeing increased deployment in epidemiology, psychology, the social sciences, and even the occasional application in physics. These techniques have, at least in how they have been commonly expressed, a close connection to the manipulationist tradition in the philosophy of causation, in so far as that the existence of a causal relation has been taken to imply a certain counterfactual conditional that involves a manipulation of, or intervention on, the purported cause. In brief, A is taken to be a cause of B in a system S only if some change in A, that was produced by an intervention, would have been accompanied by a certain change in B. I will relate this feature of the causal modeling framework especially to the theories by James Woodward (2003) and Judea Pearl (2009).
Where the manipulationist influences on causal modeling theory are present, this can lead to an interdefining of causation and intervention. I.e., causation is defined in terms of an intervention counterfactual, while an intervention in turn is given a causal definition, and this is most explicitly stated in Woodward's account. While Woodward has argued that this is not a vicious circularity, many have remained unconvinced. The circularity in definitions may appear to be a theoretical problem in its own right, but it seems to me to be especially problematic if what we desire is an intelligible naturalistic (and therefore causal) theory of what manipulations are. For this purpose, I will propose a primitivist understanding of the causal relation that occurs in the causal modeling framework. That this interpretation takes the causal relation as a theoretical primitive does not mean that there is nothing to say about it. We must still describe some properties of the relation and how these may express themselves empirically, what it is that the relation relates, and how a causal system is defined. This will thus be the topic of my talk. The hope is that this causally primitivist theory can be used to produce a causal theory of manipulations, and subsequently provide an understanding of the manipulationist account of causation as an operational definition.
This talk presents a part of my upcoming dissertation Taking Control: The
Role of Manipulation in Theories of Causation
.

References

Pearl, J. (2009). Causality: Models, Reasoning and Inference. Cambridge University Press.
Woodward, J. (2003). Making Things Happen: A Theory of Causal Explanation. Oxford University Press.