Updating, Undermining, and Independence

SPEAKER

Jonathan Weisberg

ABSTRACT

Epistemologists commonly hold that all empirical learning
is defeasible. For example, if one believes an object is red because
it appears that way, that belief may be undermined by the subsequent
discovery that the lighting conditions are deceptive. Surprisingly
though, thoroughgoing empirical defeasibility is in tension with the
standard approach to empirical learning in formal epistemology, where
empirical learning is treated as conditionalization. The problem is
not limited to probabilistic rules like classical and Jeffrey
conditionalization; it also threatens Spohn's conditionalization rules
for ranking theory and Dempster's rule for combining belief
functions. In this talk I (i) present the problem as it arises in the
probabilistic context, (ii) extend the problem to ranking theory and
Dempster-Shafer theory, and (iii) consider some proposed solutions.