Psychometric Investigations of Decomposition Methods in Subjective Probability Assessment
One approach for helping decision makers to cope with difficult decision problems is a set of techniques known as decision analysis. These techniques draw upon both normative theories of rational decision making, drawn from statistics and economics, and descriptive theories of decision making, drawn from psychology. Decision analysis relies on the general principle of problem decomposition: a large and complex problem is broken down into a representation consisting of alternatives, beliefs, and preferences. One important contribution of psychological research to decision analysis has been in understanding the methodological problems associated with the assessment of the component judgments. In particular, research has emphasized the potential for serious errors in the assessment of uncertainty. These judgments are extraordinarily sensitive to the choice of assessment procedure, and are likely to contain serious inaccuracies and inconsistencies.
The purpose of this research is to contribute to the development of an error theory that can help predict and explain the cumulative effect of assessment errors. Guided by an analytic model of error propagation in decomposition, this work will show conditions where the aggregation rule can help decision makers improve consistency despite the presence of errors. Specifically, this research involves a series of experiments investigating the effectiveness of decompositions as an aid to judging probabilities. These decompositions involve assessing the probability distribution for an uncertain quantity or event conditional upon different background events, using the apprpriate probability calculation to aggregate these assessments. The critical question is whether use of this procedure actually leads to better probability assessments. The research uses a psychometric approach to statistically estimate the extent of systematic and random error in decomposed judgments relative to judgments derived without decomposition. These experiments will explore various hypotheses about the strengths and limitations of decomposition, identifying methods for enhancing decomposition as well as circumstances where decomposition may not help.