CHI '95 ProceedingsTopIndexes
TutorialsTOC

CPM-GOMS: An Analysis Method forTasks with Parallel Activities

Bonnie E. John (1) and Wayne D. Gray (2)

(1)
Computer Science, Psychology &HCII
Carnegie Mellon University
Pittsburgh, PA, 15213, USA
Tel: +1-412-268-7182
E-mail: bej@cs.cmu.edu

(2)
Psychology & KrasnowInstitute
George Mason University
Fairfax, VA 22030-4444
Tel: +1-703-993-1340
E-mail: gray@mary.fordham.edu

© ACM

Abstract

GOMS is a family of techniques for analyzing human performance in terms of the Goals, Operators, Methods and Selection rules necessary to perform atask. Traditionally, GOMS has approximated human performance as perceptual, cognitive, and motor activities performed sequentially. However, many tasks require users to perform activities in parallel, e.g., visually searching for information, while listening to a customer, while typing. This tutorial will teach aversion of GOMS, CPM-GOMS, that predicts performance on such tasks andsaved an industrial organization millions of dollars through the evaluation of alternative system designs.

Keywords:

GOMS, user models, cognitive models, analytic methods

CONTENTS OF THE TUTORIAL

We begin with an introduction to the general concept of GOMS, that skilled human performance can be analyzed in terms of the Goals, Operators, Methods and Selection rules necessary to perform a task [1]. We examine several versions of the GOMS concept with emphasis on the relationship between the various techniques. We then present CPM-GOMS, the variant of GOMS that expresses skilled performance on tasks with parallel activities with schedule charts [2,3]. We demonstrate how to construct such a model, how to interpret its predictions, and how to make what-if evaluations of design alternatives. Attendees modify existing models and construct their own CPM- GOMS model for a new task using project management software on personal computers.

CPM-GOMS

In CPM-GOMS the parallelism of a task is represented in a schedule chart (Figure 1). Each activity in a task is represented as a box with an associated duration. Dependencies between activities are represented as lines connecting the boxes. For example, a telephone operator helping a customer cannot hit the collect-call key until s/he hears the customer request a collect call.Therefore, there would be a dependency line drawn between a box representing the perception of the word "collect" and boxes representing cognitive operators that verify the word "collect"and initiate pressing the collect-call key. The boxes and their dependency lines are drawn according to a detailed understanding of the task, goal decomposition, and operator-placement heuristics (3).

An important concept in analyzing the total task time for complex parallel tasks is the critical path. When activities occur in parallel, one sequence ofactivities will take more time than parallel sequences of activities; the critical path is the sequence of activities that takes the longest and determines the total time for the entire task. The critical path is displayed in boldface in Figure 1.

Figure 1. Critical path for task.
(Caution: Large 536857 mbyte GIF file)
.

Attendees at this tutorial learn how to create large models of a total taskfrom small building-blocks representing different activities. Figure 1 shows how the large model of a telephone operator handling a collect call (stretching across the bottom third of the page) is constructed from building block schedule charts representing system response time, the perception of sounds and visual information, conversations, hand-movements, etc.

References

1. Card, S. K., Moran, T. P., & Newell,A. (1983). The Psychology of Human-Computer Interaction. Hillsdale, New Jersey: Lawrence Erlbaum Associates.
2. Gray, W. D., John, B. E., & Atwood,M. E. (1993) "Project Ernestine: A validation of GOMS for prediction and explanation of real-world task performance." Human-Computer Interaction, 8, 3, 237-309.
3. John, B. E. (1990) Extensions ofGOMS analyses to expert performance requiring perception of dynamic visua and auditory information. In proceedings of CHI, 1990 (Seattle, Washington, April 30-May 4, 1990) ACM, New York, 107-115.