A Signal Detection Model of Compound Decision Tasks
Abstract
Detection and identification represent two fundamental types of decision tasks. Although research has focused on each in isolation, the pure forms of these tasks are generally not representative of more complex naturalistic decision environments. For example, a decision maker involved in a Search and Rescue (SAR) operation is faced with locating and identifying a crash site. This kind of decision environment is characterized by both detection and identification components. That is, the decision maker is confronted with uncertainty regarding the presence of a target crash site, and the task of identifying the target from among similar looking structures in the terrain. Decision research using compound decision tasks (detection plus identification) has the advantage of making greater contact with naturalistic environments, but carries with it the cost of increased complexity in analyzing and understanding the data. Because compound decision tasks have more than one locus where decision making can be affected, a formal method is needed to disambiguate (deconfound) effects on decision making and simplify an understanding of decision making performance in complex tasks. In this report a formal model of compound decision tasks (SDT-CD) is presented which fulfills this role. The model was assessed by an analysis of several demonstration data sets from a wide variety of content domains which highlight its ability to simplify the complexity of the task and provide readily interpretable results. In addition to measures of performance and decision bias, the model can be used to test hypotheses about decision making and permits an assessment of whether decision making is optimal.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 2006
- Accession Number
- ADA473015
Entities
People
- Matthew Duncan
Organizations
- Defence Research and Development Canada