A machine learning framework for soft skills assessment: Leveraging serious games in higher education

This study explores the use of serious games combined with machine learning techniques to evaluate student efficiency and determine appropriate degree programs. The main research question addressed is: How can predictive machine learning algorithms, applied to soft skills data collected through serious games, identify the most suitable study path for each student, thereby improving academic orientation and enhancing the likelihood of educational success? The research involved 211 university students from a single university in Italy, focusing on the application of predictive algorithms based on participants' soft skills and academic performance (GPA). The study employs logistic regression models to classify students into three degree programs: Economic Business Administration (EBA), Sports and Motor Activity Sciences (SMAS), and Modern Literature (ML). Subsequent analysis using ANOVA Kruskal-Wallis identified significant differences in predictive results across various demographic and behavioral subgroups. As an additional experiment, Data Envelopment Analysis (DEA) was employed to further understand student efficiency. DEA used soft skills as inputs and GPA as the output to calculate efficiency scores, which were then analyzed for all groups. The results highlighted significant differences in efficiency scores among different courses, age groups, and genders, but not among different technology groups or daily screen time. The findings demonstrate that serious games represent a breakthrough innovation in educational assessment, providing a highly effective platform for capturing soft skills through engaging behavioral scenarios that traditional methods cannot access. This game-based approach, when integrated with machine learning algorithms, successfully classifies students based on comprehensive competency profiles, offering a revolutionary advancement in personalized academic orientation. The serious games methodology proves superior to conventional assessment techniques by enabling real-time collection of rich behavioral data that reveals authentic student capabilities beyond academic performance metrics. Supplementary DEA analysis confirmed these findings. Despite limitations such as sample size and focus on specific devices, this study establishes serious games as a transformative educational technology that fundamentally enhances academic orientation strategies. Future research should explore larger and more diverse samples, integrate advanced technologies, and conduct longitudinal studies to further validate and enhance the serious games approach in educational contexts. • Innovative Combination of Serious Games and Machine Learning |This study integrates serious games with machine learning techniques to assess soft skills and predict academic success, offering a novel approach to educational assessment. • Predictive Models for Academic Orientation The paper applies logistic regression, ANOVA, and Data Envelopment Analysis (DEA) to predict students’ ideal academic pathways based on their GPA and soft skills, enhancing personalized education strategies. • Comprehensive Multidimensional Analysis The methodology provides a multidimensional view of student performance, combining soft skills such as time management and planning with traditional academic metrics. • Insights into the Role of Technology A detailed comparison of student performance across different technological platforms (smartphones vs desktops) reveals key insights into how device choice impacts skill development and academic outcomes. • Future Research on AR/VR in Education The paper outlines potential future applications of augmented and virtual reality in education, suggesting new avenues for immersive learning and skill assessment. • Ethical Considerations in AI-driven Assessments The study addresses critical ethical concerns related to the use of AI and machine learning for student assessment, advocating for transparency, fairness, and privacy protection.