A VR-Based Multimodal System for Assessing Presentation Integrity and Character Formation in Higher Education

Nurfaikah Nurfaikah (Universitas Muhammadiyah Makassar)
Suryadi Suryadi (State University of Jakarta)
Rachel O’Donovan (Dublin Business School)
Claire Nolan (Technological University Dublin)

Journal of Paddisengeng Technology
Vol. 2 No. 3 (2026)

Background. Artificial intelligence (AI) integration in online presentation assessment has expanded rapidly; however, existing facial emotion recognition systems in education remain largely focused on surveillance and behavioral monitoring, with limited validation against human judgment and insufficient attention to ethical and character-development dimensions. Purpose. This study aimed to develop and validate a virtual reality (VR)-based multimodal assessment framework for evaluating presentation integrity among higher education students. Method. A design science research approach was employed involving 64 students from two parallel classes (A1, n = 33; A2, n = 31) at a university in South Sulawesi, Indonesia. Data from expert validation, technical performance evaluation, NASA-TLX, System Usability Scale, and lecturer assessments were analyzed using content validity analysis with modified kappa, classification metrics, Bland-Altman agreement analysis, and descriptive statistics. Novelty. This study introduces a human-centered AI-VR assessment approach that integrates facial emotion recognition, gaze detection, and cognitive load measurement into a reflective Presentation Integrity Index. Unlike conventional AI-based monitoring systems that generate binary judgments, the proposed framework emphasizes ethical interpretation by combining affective, behavioral, and cognitive indicators to support formative character-based assessment. Results. The framework demonstrated satisfactory content validity (S-CVI/Ave = 0.83), although refinement of several indicators was required. Emotion and gaze modules achieved accuracy levels of 85.3% and 89.0%, respectively. The Presentation Integrity Index showed 93.8% agreement with lecturer assessments, moderate cognitive load (NASA-TLX = 46.7), and good usability (SUS = 76.4). Conclusion. The proposed AI-VR framework provides an ethical and reflective approach for character-based presentation assessment while requiring broader validation involving multiple evaluators and diverse educational contexts.