Abstract

Research on software application training has been predominantly based on the premise that the user builds an internal functional model of the system. This view of cognition as purely internal has been challenged by studies showing that experts do not remember the exact commands required to complete a task if they are not in front of the computer. Display-based competence suggests that a flow of information takes place between the display and the user, who only holds the necessary information to understand the visual cues and act accordingly. Distributed cognition, however, does not address the perceptual and motor actions necessary to acquire the information from the world and act on it. Embodied cognition attends to this gap. The Soft Constraints Hypothesis suggests that at the 'embodiment level', users will choose between cognitive, perceptual, and motor operators based on a cost-benefit analysis, the cost of the resources measured in milliseconds. It was hypothesized that resources in different environments with various access costs would influence use patterns, and learning strategies and performance. For the purpose of this study, participants completed a 5-lesson course in Flash animation concepts. 50 participants were then assigned to two practice environments that contained a software simulator built based on visual cues and salient task features displayed by the real software. The two practice spaces manipulated the nature of, and access to, information resources, with one featuring game-like task completion. Every action taken during the study was tracked and time-stamped, producing a log file containing around 20,000 records for subsequent analysis. The results showed that information access cost in online environments for software training has a clear impact on the strategies employed, the learning processes engaged, and learning outcomes. Traditional statistical analysis showed significant differences in declarative knowledge of rules, efficiency, and accuracy between groups. These results were complemented with data mining techniques to analyze user sequences as departures from an optimum path. The idea of a dual-learning process taking place when users are learning interactive behaviors, for example command sequences, was supported. A declarative-to-procedural process takes place in low-cost single-task environments whereas in the specific case of educational games, information access cost and concurrent game tasks can trigger a non-attentional cue-based behavior that results in higher task efficiency and accuracy, but has a negative impact on rule verbalization. A design framework for instructional environments mapping instruction-specific resources and their access cost to specific learning processes and outcomes is presented based on findings.