'Science' in the Social Sciences

7. Probabilistic Reasoning

Another response to the critics of positivism was the assertion of the central role of probabilistic reasoning and its contrast to nomological explanation.

Carl Hempel’s famous "covering law" conception of causal explanation in natural science, or the "deductive-nomological" model, held that when such explanations can be adduced and demonstrated to be true, there is a symmetry between explanation and predictive power. Thus, for example, since we know that the correct causal explanation for photosynthesis in plants is the interaction of sunlight with chlorophyll, we can predict when it takes place and when it does not. However, contrary to Hempel’s claim that such symmetry works also with "inductive-statistical" propositions, Donagan (1966) argued that probabilistic propositions are asymmetrical with respect to prediction and explanation. Thus, if I draw a white marble from an urn filled with a hundred marbles, ninety-nine of which are white and only one is black, the probability of my having done so is equal to .99. However, knowing this probability only enables one to predict with a reasonably high expectation of success which color marble could be drawn in any draw: it does not explain why I picked a white one rather than the black one. For that, a wholly different sort of story needs to be told. (For more discussion, see Coulter (1996).

Now, one can indeed argue that for some classes of explananda, having probabilistic information about their occurrence can be useful in guiding the investigator toward the ultimate goal of causal explanation, but this will not be generally the case. As Winch argued, human actions can be distinguished from purely natural events in large measure by reason of their intentionality, purposefulness, and constitution by governing rules, so that explanation of a causal, nomological sort is logically inappropriate. Nonetheless, probabilistic information about the distribution of types of activities (e.g., criminal ones of a specific type) can indeed be useful and, within limits, predictive in scope. By itself, however, a probabilistic proposition is not an explanatory one.