Pipit Utami, Yussi Anggraini, Putu Sudira, Nuryake Fajaryati, Rangga Bayu Rinawan
Affective engagement in vocational education was examined through AI based sentiment and emotion analysis situated within the Joyful, Meaningful, and Mindful learning framework. A mixed methods exploratory design integrated qualitative reflections and JMM Likert responses from 113 participants in Indonesian vocational schools (18 teachers and 95 students). Sentiment and emotion analytics were implemented in Google Colab using multilingual transformer models, with XLM RoBERTa for sentiment classification and zero shot XLM RoBERTa Large XNLI for emotion recognition, and reliability supported by double rater validation (κ =0.894). Findings indicated predominantly positive affect, with curiosity, enthusiasm, and focus as frequent emotions, while relief and frustration signalled pedagogical adaptation. Teachers showed stronger emotional regulation and self efficacy, and students contributed openness and exploratory energy. Triangulation of sentiment, emotion, and JMM components revealed affective complementarity in which stability and curiosity together supported meaningful engagement. The study advances emotion aware analytics in vocational learning by offering a reproducible and human centered template for mapping affective depth in TVET pedagogy. © 2025 IEEE.
Universitas Negeri Yogyakarta, Department of Electronics and Informatics Engineering Education, Sleman, Indonesia; Universitas Negeri Yogyakarta, Study Program of Technology and Vocational Education, Sleman, Indonesia; Universitas Negeri Yogyakarta, Distance Education Program in Technology and Vocational Education, Sleman, Indonesia
