Building Predictive Human Performance Models of Skill Acquisition in a Data Entry Task

Abstract

This paper presents a predictive model of a simple, but important, data entry task. The task requires participants to perceive and encode information on the screen, locate the corresponding keys for the information on different layouts of the keyboard, and enter the information. Since data entry is a central component in most human-machine interaction, a predictive model of performance will provide useful information that informs interface design and effectiveness of training. We created a cognitive model of the data entry task based on the ACT-R 5.0 architecture. The same model provided good fits to three existing data sets, which demonstrated the effects of fatigue with prolonged work, repetition priming, depth of processing, and the suppression of subvocal rehearsal. The model also makes predictions on how performance deteriorates with different delays after training, how different amounts of rehearsal during training affect retention, and how re-training helps retention of skills.