Analysis of Machine Learning-Based Investigation Into Multivariate Factors of Team Performance in Serious Games:...

Area of Science:

• Behavioral Science
• Computer Science
• Organizational Psychology

Background:

• Serious games (SGs) are increasingly used to study team performance.
• Prior research often isolates leadership and communication, neglecting multivariate interactions.
• Understanding these interactions is key for improving training and research.

Purpose of the Study:

• Develop machine learning (ML) models to predict team success in SGs.
• Identify key behavioral and demographic predictors of team performance.

Main Methods:

• Retrospective analysis of 233 teams in escape room SGs.
• Data collected on collaboration, communication, and leadership via trained observers.
• Four ML models (logistic regression, random forest, MLP, SVC) trained and evaluated using 5-fold cross-validation.

Main Results:

• Winning teams showed higher scores in knowledge sharing, leadership, guidance, and extraversion.
• Logistic regression model achieved the highest accuracy (88%).
• SHAP analysis identified gender composition and prior experience as key demographic predictors, and specific behaviors like 'celebrating progress' as crucial behavioral indicators.

Conclusions:

• Multivariate analysis in SGs offers deeper insights than isolated factor studies.
• ML models accurately predict team success based on behavioral and demographic data.
• Findings can inform strategies to enhance team productivity in organizational settings.