Christian S. Loh, PhD, Director, Virtual Environment Lab, Southern Illinois University Carbondale I-Hung Li, Research Assistant, Virtual Environment Lab, Southern Illinois University Carbondale
Serious games were originally meant to become advanced training tools to improve decision-making skills and raise job performance in trainees/learners. Currently, less than 10% of serious games have been designed to facilitate training – including those used for military and medical simulation and training. Serious games analytics can increase the value of these types of training if we know what performance metrics to use, how to turn these analytics to meaningful insights for stakeholders, as well as provide just-in-time (re)training and remediation for the learners, which can impact Return of Investment (ROI) for the learning organizations.
The behavioral and cognitive differences between skilled individuals (experts) and novices have been well-documented in the literature. Action sequence, a component in competency, comprised of the chronological order of actions performed, has been shown to be useful in differentiating (likely) experts from novices for training performance assessment purposes. The ability to discriminate experts from novices and possibly to predict who might become experts faster (shorter period of training) is highly desirable in the training and learning industries.
We captured the actions of players in a serious games and coded their action-sequences by dividing the game world into grids of different granularity. We use the Partial Least Squares Discriminant Analysis (PLS DA) function in JMP Pro 12 to compare these variables, along with total game completion time, to differentiate the players into experts and novices based on their behaviors in the game for performance assessment and serious games analytics.
Keywords: Partial Least Squares Discriminant Analysis, Serious Games Analytics, action sequences, similarity index, performance assessment