A structured analysis and quantitative measurement of task complexity in human-computer interaction

Baijun Zhao, Purdue University


Task complexity is an important factor that must be considered in the design of human-computer systems in order to reduce human workload (especially mental load) and enhance human task performance. Traditional measurement techniques of task complexity are based on subjective and qualitative judgments. In this study, a structured task analysis methodology together with a quantitative measurement of task complexity is proposed. With respect to the theories and empirical findings on memory structure and coding, attention resources, human information processing and knowledge representation, a conceptual model is developed which defines three dimensions of task complexity: task size, task characteristic and task presentation. It is hypothesized that (1) a logarithmic relationship exists between task complexity and task size; (2) there is a significant interaction between task presentations and task characteristics, and the increased compatibility of task presentations and task characteristics reduces task complexity. In order to measure task complexity quantitatively, a practical methodology of structured task analysis (STAM) was developed. This method decomposes a task systematically into the unit components which can be directly measured on the three dimensions of task complexity. A nested factorial experiment was designed and carried out to test the hypotheses. Thirty-two subjects were asked to perform tasks with different task sizes and different degrees of task presentation-characteristic compatibility. Subjects performance data (completion time and error rate) as well as subjective ratings of mental workload and task difficulty were collected. Although the logarithm model fits significantly the experimental data by the regression and correlation analyses, hypothesis one is not fully supported because there is no statistically significant difference between the linear model and logarithm model. Therefore, both the linear model and the logarithm model can well predict task complexity. The ANOVA results supported hypothesis two. There is a significant interaction between task presentations and task characteristics such that the flowchart technique is more effective than the language-like presentation in performing conditional branching procedure tasks.




Salvendy, Purdue University.

Subject Area

Industrial engineering|Computer science

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server