Unification and description-variance in the philosophy of science.

摘要

Philosophers and scientists have powerful intuitions that unification has evidential force. A hypothesis that unifies a body of propositions taken for granted is better confirmed. But why should this be so? We want to know (1) what unification is, and (2) why it would be evidentially virtuous. This is the problem of unification. It cannot be solved without specifying what it means for evidence to confirm a hypothesis. In this dissertation I select one broad theory of evidence and show that given this theory, unification cannot be evidentially virtuous as long as unification understood in one of three ways.;The broad structure of the dissertation is as follows. First I discuss the broad theory of evidence that I take for granted. I call it 'logicism', and contrast it with a rival theory of evidence called 'historicism'. Second, I consider two styles of attempted solutions to the problem of unification: one is the idea that the unification of a body of propositions by a hypothesis consists in the body's having a boost in its coherence conditional on the hypothesis, and the other says that unification is predicated on the notion of varied or diverse domains of evidence. I demonstrate that both of these are beset with a dependence on how propositions are pictured rather than the propositions themselves. Third, I consider the import of this description-variance, and argue that it is fatal to the view that these two types of unification are evidentially virtuous. Fourth, I consider one more strategy for solving the problem of unification, namely the view that unification is related to constraints among a hypothesis's parameters; this is a bigger project because I first try to get a grip on some fundamental questions such as 'what is a parameter?'. In the end, this approach appears to suffer a similar fate.