Dynamic Signal Detection Theory : The next logical step in the evolution of signal detection analysis

Measurement, design, and training programs in human factors often rely on the formal concepts of signal detection theory in order to analyze and interpret the performance of human operators. Following recent work in the response time modeling literature, we propose a generalization of this classical framework that is as simple to apply and yet more powerful, in the sense that it is able to identify speed-accuracy trade-off effects that would be confused by detection theory with sensitivity effects. Like the classical, static detection models, this dynamic framework distinguishes between the sensitivity of the operator and operator biases that may lead to higher identification rates for some events but at the expense of lower identification rates of other events. We deve lop formulas for the three free parameters of the dynamic model that can be computed easily without requiring a parameter search routine. We also show how two different kinds of biases, in the stopping rule and in the decision rule, can be distinguished and related to conditions of the display or to operator preferences. These methods require no additional design features and make it possible for investigators not only to quantify an operator’s ability to perform a discrimination task but also to provide feedback that operators can use to correct any suboptimality in their decision making strategy.