Empirical Evaluation of a Decision-Analytic Aid.
Abstract
An experiment was performed to investigate the potential of DDI's Rapid Screening of Options (OPINT) minicomputer-based aiding package for enhancing decision making. Two different politico-military scenario backgrounds and message sets were developed (familiar and unfamiliar), and each was further elaborated and configured into a set of five intelligence summaries (i.e., scenario version), which cumulatively reflected an enemy intention of either 'attack' or 'no attack.' The ground truth for each scenario version was then validated in a preliminary empirical study so that it could serve as a criterion for evaluating performance in the subsequent main experiment. The subjects in the main experiment were 24 experienced naval intelligence analysts, with 12 participating in an aided experimental condition and 12 in an unaided control condition. Each analyst was given one version of the familiar scenario and the opposite version of the unfamiliar scenario. After reading each intelligence summary, he was required to diagnose the enemy's military intentions and to make a decision by recommending one of four prespecified courses of action. The use of the aiding package significantly increased the number of correct decisions under the attack version of the scenarios, but not under the no-attack version; this result was attributed to content differences between the scenarios as well as to a conservative utility function on the part of the analysts.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1980
- Accession Number
- ADA087886
Entities
People
- Francois G. Christen
- Michael G. Samet