Abstract
This chapter discusses the basic principles of Bayesian inference and some epistemological issues that emerge from this formalism. It highlights some of the difficulties associated with using the logical approach instead of the probabilistic approach. The chapter discusses the difficulties associated with the nonmodularity of plausible inferences, that is, the impropriety of drawing conclusions from certain truths in the database without checking other truths that may reside there. It also discusses the problem of query, sensitivity, which stems not from neglecting facts that were learned but from neglecting to specify the facts that could have been learned. Formalisms that ignore the query process altogether are bound to be insensitive to an important component of human reasoning. The treatment of virtual evidence, using the vector of likelihood ratios, sidesteps the requirement of specifying a full protocol in advance. This option expands the repertoire of Bayes analysis by permitting the assimilation of evidence by means other than straight conditioning, and it simultaneously facilitates the manipulation of belief updates within the traditional syntax of probability calculus.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have