Feedback in a complex multiman-machine system

Abstract

DECISION MAKERS PROVIDED HYPOTHESES AND MADE DIAGNOSES, IN THE FORM OF CONDITIONAL PROBABILITY JUDGMENTS, TO ACCOUNT FOR THE OCCURRENCE OF CERTAIN CRITICAL EVENTS IN A SIMULATED HOSTILE ENVIRONMENT. THE DECISION MAKERS' ESTIMATES WERE COMPARED WITH SIMILAR ESTIMATES PROVIDED BY A BAYESIAN MODEL FOR SEVERAL LEVELS OF PERCENTAGE OF KNOWLEDGE OF RESULTS (KOR) AND 2 LEVELS OF SPECIFICITY OF KOR (ACCESS OR NO ACCESS TO MODEL ESTIMATES). THERE WERE NO SIGNIFICANT DIFFERENCES IN THE PROBABILISTIC ESTIMATES PROVIDED FOR 33%, 67%, AND 100% KOR BUT ALL 3 WERE SUPERIOR TO 0% KOR. DECISION MAKERS WITH ACCESS TO BAYESIAN MODEL ESTIMATES AS FEEDBACK WERE NOT ABLE TO IMPROVE THEIR JUDGMENTS SIGNIFICANTLY EVEN THOUGH THE MODEL-GENERATED SOLUTIONS WERE SIGNIFICANTLY SUPERIOR TO HUMAN ESTIMATES AT ALL KOR LEVELS ABOVE 0%. (PsycInfo Database Record (c) 2025 APA, all rights reserved)