This project aims to use an account of evidence based on error-probabilities as the starting point for developments toward a broader-reaching epistemology of science. It is hoped that the work will constitute progress toward a satisfactory integration of perspectives from philosophical approaches to epistemology, methodological studies of theory confirmation, theories of statistical inference, and the study of science as an inherently social enterprise.
The project draws upon Deborah Mayo's error-statistical account of scientific evidence, according to which evidence accrues to hypotheses that pass severe tests, i.e., when testing procedures yield data that fit the hypothesis in question, and are such that it is highly improbable that the hypothesis would have passed a test with data fitting as well or better, were the hypothesis false. Although this account is methodologically fruitful, its epistemological import has not been made fully explicit.
The first part of the project outlines some elements of an error-statistical epistemology of science that supplements Mayo's account with attention to the role of abductive inference, the concept of the security of inference, and a perspective from argumentation theory. More specifically: the project considers a way to supplement the error-statistical account of induction in a way that shows the role necessarily played by abductive inference in the justification of experimental models; the concept of secure inference is defined and defended, as distinct from considerations of reliability, as an important component of epistemic evaluation in science; and insights from argumentation theory are deployed to show how standards of argumentative cogency might be developed within an error-statistical perspective.
Part two explores some applications of the framework developed in part one, to show how it can address a gap between the kind of detailed study of experimental practice that motivates the work of the "New Experimentalists," who have advocated an approach to philosophy of science emphasizing experimental practice, and the evaluation of very general scientific theories. Examples from physics are used to show how piecemeal tests can provide a basis for the assessment of such very general theories as General Relativity and Quantum Electrodynamics, and how a concern for the security of inference sheds light on the use of experimentally testable physical theories to solve problems of "fine-tuning" (by contrast to the theological responses that fine-tuning has inspired among some philosophers and theologians).
The project draws upon recent work in the error-statistical approach to scientific inquiry that has attracted the interest of scientists, statisticians, and philosophers of science. It aims to advance the error-statistical approach by clarifying its philosophical import. The intended benefit for students of scientific inquiry is a clearer understanding of how experimental science generates knowledge. In addition, the project is directed at clarifying and drawing attention to security of inference, an important but philosophically neglected aspect of methodological appraisal.
It is hoped that this project will result in a published book that will be of interest to working scientists as well as to philosophers, statisticians, and other students of scientific inquiry. Many philosophers outside of the specialty of philosophy of science, including epistemologists, have a very limited perspective on our current understanding of scientific evidence. By explicitly addressing core epistemological concerns, this work may bring the attention of the broader philosophical community to the rich methodological perspective offered by approaches deriving from statistics. By applying the philosophical framework to specific problem-contexts in the physical sciences, it is hoped that the work will be shown to have relevance beyond the philosophy seminar room.
