Toward an Error Theory for PIP (Probabilistic Information Processing): Inference Based on an Alternative Formulation of the Data Space
MICHIGAN UNIV ANN ARBOR ENGINEERING PSYCHOLOGY LAB
Pagination or Media Count:
Probabilistic Information Processing PIP systems, as currently conceived, use experts intuitive judgments about the diagnostic impact of individual data as inputs for mechanical aggregation by Bayess theorem. Past research has shown that the posterior odds output by PIP are much more extreme than those arrived at via human aggregation. Because of this superior efficiency PIP-type processing of fallible data has been recommended as an important tool for decision making. The present paper questions the uncritical use in PIP of estimated likelihood ratios as if they were veridical. A theory is developed which incorporates into the inferential process the inherent variability of human judgment. The resulting effect is a decrease in the posterior odds given by PIP.