Against Conditional Probability
Abstract
While a minority of statisticians hew to the Bayesian line, a large number of philosophers and a large number of AI researchers take Bayesian conditionalization for granted as the only way of updating uncertainties. At the same time, everybody, Bayesian or not, appears to accept the fundamental principle of direct inference: if you know the statistics, the statistics should constrain your belief. The contribution of this paper is to exhibit a conflict between these two principles, and to argue in favor of direct inference and against conditionalization.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1989
- Accession Number
- ADA250344
Entities
People
- Henry E. Kyburg Jr.
Organizations
- University of Rochester