Science has achieved such great success that we are tempted to believe scientific theories are true or at least approximately true. Scientific anti-realists, however, argue that empirical success is not a good indicator of a theory’s truth because a lot of theories that were once successful later turned out to be false.

One way of responding to their “pessimistic induction” is to claim that some of false-but-successful theories were still partially true and thereby successful. Scientific realists hope to discover that if past scientists had believed in the truth of only the components of past theories that were responsible for their success, they could have evaded the challenge of the pessimistic induction. Versions of scientific realism which selectively determine credible parts of successful theories are generally called “selective realism.”

But how could selective realists identify which components of a past theory were responsible for its success? If they judge certain aspects as irrelevant just because those parts are no longer preserved in present theories, realists are guaranteed to find historical patterns of theory-change that are favorable to their views. Discussants on both sides, therefore, assume that realists have to propose prospectively applicable criteria for identifying working and idle parts of a theory without taking advantage of having a retrospective viewpoint.

Selective realists attempt to propose criteria for selecting credible aspects of a theory and to show that past scientists could have evaded the challenge of the pessimistic induction had they applied the criteria proposed by selective realists instead of the ones they actually applied. Note that this amounts to proposing rules of confirmation or induction and to showing that their rules are better than the ones applied by past scientists. But how could we justify the goodness of given inductive rules? It seems that the best empirical support for inductive rules, upon which both realists and anti-realists could agree, is the fact that the rules have not failed so far. Realists cannot abandon retrospective judgment because only from the present viewpoint, could they justify the superiority of the inductive rules they recommend over the ones past scientist actually used. We cannot show that one inductive rule is superior to another insofar as both fit the collected data. Realists need not, however, abandon retrospective judgment because if they are asked to justify their belief, they should refer to all relevant available evidence, which is the evidence available at present, not the evidence available at some past point.

(Satoshi Kudo)