Affective learning and the charismatic lecturer

Students can be inspired, engaged and motivated by lecturers who possess the elusive property of charisma. This study harnessed affective computing to map students' perception into quantitative voice features of lecturers. A cohort of 900 students listened to 200 short audio extracts from academic lecture recordings. The students provided rating of the lecturers' charisma using two validated scales. Explainable prosodic voice features were extracted from the lecturers' speech recordings. A random forest classifier was trained on the lecturers' voice features and students rating labels. The features were ranked by the classifier according to their contribution to the prediction F1-score. Six out of the ten upper-ranked features were rhythm change features. The prediction accuracy exceeded 0.9 for all charisma rating classes. The results portray the potential of affective computing to explain the property of charisma in lecturers' voice. Our findings highlight features in the production and perception of charismatic voice of lecturers. This affective model can potentially be used for machine-guided practice for lecturers and improve their teaching effectivity and their interaction with learners.