Objectivity in confirmation: Post hoc monsters and novel predictions
Introduction
The study of confirmation is the study of the conditions under which a piece of evidence supports, or ought to support, a hypothesis as well as of the level of that support. There are two major kinds of confirmation theories, objective and subjective. Objective theories hold that confirmation questions are settled by purely objective considerations. Subjective ones hold that at least some non-objective considerations come into play. With some exceptions (see, for example, Williamson, 2010), most confirmation theorists nowadays opt for subjective theories. The pessimism over objective theories is most probably due to the fact that it has proved very hard, some may even say impossible, to find reasonable principles that decide questions about confirmation in purely objective terms. The aim of this paper is to reverse some of that pessimism by putting in place some cornerstones in the foundations for an objective theory of confirmation. This is achieved by considering lessons not from the failures of subjective theories, which, no doubt, there are many, but rather from the failures of a certain kind of mini-theory of confirmation, namely predictivism, that is typically conceived of as objective.
Section snippets
The completion challenge
Imagine a scientist S who endorses a hypothesis H but is confronted with incontrovertible evidence E that contradicts H. Unless S is a defeatist, two options seem to be available. S can attempt to either modify H or else construct an entirely new hypothesis. Whatever the chosen option, S will have to ensure that the hypothesis endorsed stands in the right kind of inferential–semantical relations, ideally the entailment of true propositions, to the wayward and other established empirical
Predictivism
The requirement that, in addition to standing in the right kind of inferential–semantical relations to the evidence, hypotheses make predictions is a way to meet the completion challenge, either in part or in full. According to this movement, accommodated evidence and, for some of its theorists, even evidence that could have been accommodated, i.e. evidence that could have been used in post hoc constructions, is somehow inferior to predicted evidence, or, as such predictions are sometimes
Temporal novelty
One particular brand of predictivism is temporal novelty. According to this view, support emanates solely (or more plentifully) from phenomena that become known after the hypothesis, plus any auxiliaries, that predicts them was formulated or modified.5 That is, it emanates from phenomena that are temporally novel. This puts the view squarely in the objective theory of
Use-novelty
At least some data that were known prior to the formulation or modification of a given hypothesis, it seems, must be capable of providing support (or as much support) for that hypothesis.
The complete data set counterexample
In this section, I would like to briefly discuss an important counter-example to Worrall’s theory and use-novelty more generally. Mayo (1996, p. 271) asks us to imagine a logic class which contains students who took the SAT examination. Suppose we want to find out the average SAT score of those students. Suppose further that we have access to all the scores. The best way to approach the matter is to add up all the individual scores and divide the resulting number by the number of students in
Contingency and conflicting confirmational assessments
Unlike Zahar (1973), Worrall does not urge us to peer into the notebooks of scientists in order to assess whether data has been used to construct a theory or auxiliary. He is wholeheartedly bent on eliminating psychologically contingent considerations from his account: “My account gives no role to any such psychological factor” (2005, p. 819). And it is not only psychological factors that he dismisses. As we saw in Section 4, he laughs off the idea that any “contingent historical issue”
A pandemic for incidental predictivists
Worrall’s view is the only brand of predictivism that attempts, and as I just argued fails, to exclude contingent considerations from confirmational matters. All other brands of predictivism deliberately include such considerations in their calculations. Let us call these predictivist views “incidental.” As we saw earlier, the inclusion of contingent considerations is not in and of itself objectionable, for such an objection would beg the question against incidental predictivists. Even so, I
Cornerstones for an objective theory of confirmation
It must seem like ages since we last spoke about objectivity and subjectivity in confirmation. Yet a highly pertinent discussion has already been taking place right under our noses. The failures of predictivism exposed in the last three sections are highly instructive in our search to lay some foundations for an objective theory of confirmation. In this section and the next, I would like to place some cornerstones in these foundations in the form of four desiderata that an adequate objective
Post hoc and other monsters
No discussion of desiderata for an objective theory of confirmation would be complete without saying something about the much-derided post hoc monsters. Recollect that to meet the completion challenge requires, among other things, dealing with the niggling issue of whether or not, and if so to what extent, post hoc monsters can be confirmed. One approach to attain that goal has been to either demonise or penalise post hoc-ness itself. This approach has been taken up by the predictivist
Acknowledgements
I have benefitted from discussions with, and would therefore like to thank, José Díez, Ludwig Fahrbach, Clark Glymour, Carl Hoefer, Stathis Psillos, Sam Schindler, Gerhard Schurz, Paul Thorn, and John Worrall. I am also thankful to two anonymous referees. Referee 2, in particular, provided exceptionally incisive comments. I gratefully acknowledge the German Research Foundation (Deutsche Forschungsgemeinschaft) for funding my research under project B4 of Collaborative Research Centre 991: The
References (23)
Changes in the problem of inductive logic
Prediction and the ‘periodic law’: A rejoinder to Barnes
Studies in History and Philosophy of Science
(2005)The paradox of predictivism
(2008)The aim and structure of physical theory
(1991)- et al.
Einstein and Hilbert: Two months in the history of general relativity
Archive for History of Exact Sciences
(1978) - Glymour, C. (2008). The paradox of predictivism (book review). Notre Dame philosophical reviews,...
Fact, fiction and forecast
(1983)On the predilections for predictions
British Journal for the Philosophy of Science
(2008)- et al.
Prediction versus accommodation and the risk of overfitting
British Journal for the Philosophy of Science
(2004) The apparent superiority of prediction to accommodation as a side effect: A reply to Maher
British Journal for the Philosophy of Science
(2001)
Empirical equivalence and underdetermination
Journal of Philosophy
Cited by (5)
Prediction in epidemiology and medicine
2015, Studies in History and Philosophy of Science Part C :Studies in History and Philosophy of Biological and Biomedical SciencesWhat Science Fiction Can Demonstrate About Novelty in the Context of Discovery and Scientific Creativity
2019, Foundations of ScienceDerivational Robustness and Indirect Confirmation
2018, ErkenntnisAd Hoc Hypotheses and the Monsters Within
2016, Synthese LibraryUnification: Not just a thing of beauty
2015, Theoria (Spain)