Convention, Confirmation, and Credibility

Abstract

Abstract : There are good arguments for conventionalism in science, but that if they are pursued too far they leave us with empty bodies of knowledge. We must have grounds for preferring one convention to another, and furthermore, they must be epistemic grounds. Such grounds cannot be provided directly by observation. As many conventionalists have pointed out, observation is always subject to error. The notion of error, and hence of observationality, is relativized to a theory, but is determined, for a given theory, by our experience - by what happens to us. Thus even though theories are regarded as 'conventional' or a priori, even though the notion of error that determines the content of the practical certainties of a theory is internal to the language of the theory, we can still have an objective measure of the degree to which one theory rather than another satisfies our desire to anticipate the future. This picture leaves certain puzzles to be resolved, but seems to offer a picture of scientific knowledge that is plausibly rooted in the empiricist tradition. It may be that the general framework of metalinguistic report at one extreme, and object language prediction at the other.