JMIR Publications

ABSTRACT

Background:

Various organizations increasingly use serious games to study team behavior with an aim to improve team performance. Previous studies in this regard were mostly focused on studying leadership and communication related aspects on an individual basis. However, the complex multivariate relationships between various behavioural indicators remain unexplored, which can potentially provide deeper insights into the factors, which actually determine team performance. An understanding of these multivariate factors will contribute to the development of better teams, resulting in improved performance and productivity in workspaces.

Methods:

This study employs a machine learning based predictive modelling approach to identify influential behavioural and demographic predictors that lead to team success. A dataset of 233 teams was first analyzed using exploratory analysis techniques, and then four different types of machine learning models, namely Logistic Regression (LR), Random Forest (RF), Multi-layer Perceptron (MLP) and Support Vector Classifier (SVC) were developed to classify between the winning and the losing teams. The best performing model in terms of balanced accuracy score was analyzed using suitable explainability methods to identify the most significant features related to various aspects of behaviour that drive the outcomes of the model and consequently for the game.

Results:

Exploratory data analysis revealed that winning teams consistently outperformed losing teams across constructs such as environmental awareness, leadership and guidance, and displayed differentiated communication patterns. The overall accuracy of all developed machine learning models was above 79%, and the best-performing model exhibited an accuracy of 81%. The application of explainability methods on the best-performing model showed positive reinforcement behaviours and adaptive leadership actions as the strongest predictors of team success.

Conclusions:

This study shows that exuberant behaviours, adaptive leadership, guidance and monitoring and gender composition consistently surfaced as the strongest predictors of team performance in serious gaming.