Improving interface quality: an investigation of human-computer interaction task learning

Abstract

User learning is of critical importance in evaluating interface usability (and in turn interface quality). The focus of this research in on interface leamability, where a stochastic model represents the learning process required for successful completion of human-computer interaction tasks. The parameter used to quantify learning is a learning rale. Of interest here is the validation of learning rate as a measure of interface quality. Learning rate was validated against two traditional measures of interface quality: task completion time, and error frequency. SuperCard(tm), a Macintosh project utility, provided an empirical learning environment in which 32 participants learned 16 fundamental SuperCard tasks. Results of correlation analyses suggested the usefulness of learning rate as an indicator of interfacequality. Our learning rate analysis identified four tasks presenting learning difficulties. (Analysis of task completion times identified two of these four tasks, and error frequency analysis identified one). Learning rate data captured all of the information available from the two traditional interface quality measures and identified two tasks disregarded by them. Incorporating learning rates in the interface evaluation process precludes time-intensive video tape analysis typically required by more traditional interface quality measures.

Keywords:

• Human-computer
• interaction
• Interface
• usability
• Interface
• evaluation
• Learning
• models