A Two-Level System of Knowledge Representation Based on Epistemic Probability

Abstract

A knowledge state is represented by two sets of statements, rather than one. One set of statements represents evidence; it corresponds to recorded data, together with general knowledge that is not open to question in the context at hand. We refer to this as the evidential corpus of knowledge. The other set of statements represents a body of practical certainties, based on the statements constituting the evidential corpus. It consists of statements whose probabilities, relative to the evidential corpus, exceed some explicit level determined by the context. Probabilities are assigned to statements, relative to a body of evidence called evidential corpus. We require statistical knowledge (not just statistical evidence) as a basis for every probability statement. Two facts render this constraint acceptable: It doesn't take much statistical data to yield an approximate statistical hypothesis. And if we adopt the principle that statements known to have the same truth value are to be assigned the same probability, we may link many statements to the same statistical foundation.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 01, 1989
Accession Number
ADA255280

Entities

People

  • Henry E. Kyburg Jr.

Organizations

  • University of Rochester

Tags

Communities of Interest

  • Biomedical

DTIC Thesaurus Topics

  • Artificial Intelligence
  • Bayesian Networks
  • Data Science
  • Expert Systems
  • Formal Languages
  • Information Science
  • Intelligent Systems
  • Intervals
  • Language
  • Logic
  • Optical Scanning
  • Philosophy
  • Probability
  • Reasoning
  • Statistical Data
  • Statistical Inference
  • Uncertainty

Readers

  • Adaptive Control and Estimation with Uncertainty in Dynamic Systems.
  • Instructional Design and Training Evaluation.