A STUDY IN PROBABILISTIC INFORMATION PROCESSING (PIP)
Abstract
Men required to choose among alternative hypotheses given fallible data fail to extract as much "certainty" as the data justify. PIP Theory indicates that performance can be improved by using Bayesian probability judgements. This experiment was designed to test this theory. Subjects, receiving simulated military data, determinined which one of three strategies (Hypotheses) an enemy was using. In the NON-PIP condition subjects estimated P(H/D) directly. In the PIP condition subjects estimated P(D/H), and P(H/D) was calculated using Bayes Formula. Results show: (1) the highest probability was always assigned to the correct hypothesis. (2) PIP was superior to NON-PIP in (a) achieving higher posterior probabilities, and (b) reaching asymptotes faster. (3) Increasing difficulty resulted in poorer performance in NON-PIP.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 02, 1963
- Accession Number
- AD0402145
Entities
People
- J. R. Newman
- Richard J. Kaplan
Organizations
- System Development Corporation