The Predictability of Cursor Control Device Performance Based on a Primitive Set of User Object-Oriented Cursor Actions

Abstract

The ability to predict performance with a cursor control device on a complex task by measuring performance on a simple task would be useful in evaluating alternative input devices in many types of novel situations. A user would simply have to perform simple cursor movements with each candidate device, and predictions could be made of his/her performance with the devices on any given software application. Such an approach would reduce tedious trial and error procedures, as well as eliminate the time necessary to first learn various software applications. The current study employed the User Action Notation (UAN), a task-oriented notation that describes the behavior of the user and the interface during their cooperative performance of a task, to decompose complex tasks into primitive components. A set of primitive cursor actions was developed which contains the elementary cursor actions found in complex tasks. A graphics software application was then evaluated, using the UAN, with respect to the frequency of occurrence of each of the primitive user-cursor actions. Individual's ability to perform each primitive user-cursor action with three different input devices was then be measured. These measures were used to form estimates of the individual's ability to perform the graphics task with each input device. Correlations between predicted performance and measured performance on the graphics task were found to exceed 0.9. Results demonstrate the success of the method described herein for predicting complex task performance based on simple task performance, as well as, the usefulness of the UAN for decomposing complex tasks into primitive components.