Abstract:

David Hume appropriated skeptical arguments, first formulated in Plato’s Meno and later by Sextus Empiricus, to argue that causal claims cannot yield predictions sure to be true in every logically possible world. The (sound) argument in Plato and Sextus applies to non-deductive inference in general, and it was purely an accident of the times that Hume wrote about causality, but, ever since, causality has been the almost always indispensable, and almost always suspect, scientific concept.

This lecture will provide a motorcycle tour of some of the history of thinking--from Francis Bacon to Pierre de Laplace to John Stuart Mill to George Boole to Charles Peirce to Ronald Fisher to Carl Hempel--about how causal relations can and cannot be estimated from phenomena, culminating in the mantra taught in every introductory methodology class: correlation is not causation. Then I will turn to survey some of the dramatic, philosophically driven, developments of the last thirty years in methods of causal inference, developments that should have repressed the mantra and altered the practice of many scientific disciplines. I will conclude with some speculations about why so little of that practice has changed in consequence.