Epistemological Relevance and Statistical Knowledge
This is an incremental analysis of a foundational problem in AI and epistemology, with no new solutions proposed.
The paper addresses the long-standing issue of lacking adequate statistical knowledge for managing epistemic uncertainty in belief and inference, highlighting that definitions and applications of such knowledge remain unspecified.
For many years, at least since McCarthy and Hayes (1969), writers have lamented, and attempted to compensate for, the alleged fact that we often do not have adequate statistical knowledge for governing the uncertainty of belief, for making uncertain inferences, and the like. It is hardly ever spelled out what "adequate statistical knowledge" would be, if we had it, and how adequate statistical knowledge could be used to control and regulate epistemic uncertainty.