Published online by Cambridge University Press: 28 February 2022
Rather than distinguishing between the natural and social sciences on the basis of their subject matter, it is often philosophically fruitful to start with the fact that most social phenomena are not amenable to investigation by laboratory experimentation. One can, in some cases, use randomized experimental designs to investigate causal relationships between social phenomena, but ethical and practical difficulties often preclude the use of such techniques. In the absence of these two empirical approaches, two principal alternatives are often used: causal models and quasi-experimentation. I shall focus on the former here, because the techniques involved offer an interesting example of how the context of investigation can affect the structural form of mathematical models used to describe phenomena. I have an ulterior motive, however. Little attempt has been made to link the substantive work on causal modeling with philosophical theories of probabilistic causality.