Trout 50 Years of Successful Predictive Modeling Should Be Enough: Lessons for Philosophy of Science

Citation: Michael A. Bishop and J. D. (2002) Trout 50 Years of Successful Predictive Modeling Should Be Enough: Lessons for Philosophy of Science. Philosophy of Science 69 (RSS)
Internet Archive Scholar (search for fulltext): Trout 50 Years of Successful Predictive Modeling Should Be Enough: Lessons for Philosophy of Science
Download: http://www.jstor.org/stable/pdf/10.1086/341846
Tagged:

Summary

The main purpose of this article is to highlight importance of predictive modeling relating some questions in the philosophy of science. Especially, the authors focus on the question of statistical prediction rules (SPRs) in this article because the success of SPRs give us better understanding, explanation, reasoning and the way how we should do philosophy of science.

SPRs are one of actuarial models which provide a purely mechanical process for arriving at predictions on the basis of quantitatively coded cues. And the predictions of SPRs are much better than the predictions of human experts (the Golden Rule of Predictive Modeling) when based on the same evidence. But there are some limitations of SPRs because it is hard to apply the SPRs into human and social prediction. However, there is also solution called the flat maximum principle which has broad implication on the basis of the fact that human and social prediction can be explained in the similar vein and on extension of similar mechanism of others.

There are some problems of SPR. We don’t know the linkage between frequent probability claim and a judgment about particular case and don’t know how to deal with error (“broken leg”). Those problems are related with our overconfidence in our judgment and its bias goes far beyond our self-assessments (Even well trained scientists cannot be free of the bias). However, the overconfidence also has benefit that it can defend premature but promising new idea against traditional one.

In terms of the authors’ perspective about scientific explanation, epistemological approach is reasonable. Scientific explanation also has limitation of precision between scientific account and natural phenomenon. The authors think that the problem can be solved by case studies, but it also contains limitation of validity of several case studies because we can never apply all cases. And that limitation becomes the reason why actuarial approach is needed.