Causally Licensed Inference and the Confirmation Relation
Dissertation, University of Pittsburgh (
1991)
Copy
BIBTEX
Abstract
This dissertation proposes a new sort of explication of the relation between a hypothesis and the data which test it. Instead of deductive, explanatory, or Bayesian probability relations, qualitative confirmation is explicated in terms of causal relations--E is evidence of H if and only if E is a causal consequence of H. Evidential inferences, according to this confirmation theory, are to be restricted to those licensed by underlying ontic relations. I show how the inferential properties of the causal relation match intuitions about evidential relevance and thereby avoid the paradoxes of confirmation, including the Grue and Raven Paradoxes. This approach meshes with Salmon's ontic theory of scientific explanation, thus providing a unified integration of confirmation and explanation, and revealing the source of the warrant for the "inference to the best explanation" while additionally showing how that theory is incomplete. I also criticize Bayesianism and the hypothetico-deductive method, showing how the causal view's explication of evidential relevance is superior. Applications of this "Hypothetico-CausoInferential Method" are given using several historical scientific examples, including Oldham's discovery of the internal structure of the earth, Darwin's theory of evolution, Einstein's light-quanta hypothesis, and current debate on what killed the dinosaurs. The theory is compared and contrasted with similar views found in Aristotle, Descartes and Miller. It is judged against criteria of adequacy suggested by Hempel and Glymour, and I argue that it provides the best explanation of why controlled experimentation is taken as the hallmark of the scientific method. Finally, a strong version of the theory is considered which holds that in ideal circumstances a particular hypothesis may be confirmed with practical certainty by a single experiment, as Hume suggested. This possible notion of "inductive validity" is based upon a productive necessity-and-sufficiency analysis of causation, which I defend in detail against various objections using a proposed three-place model of the causal relation.