Cognitive Benefits of Employing Multiple AI Voices as Specialist Virtual Tutors in a Multimedia Learning Environment

Limited prior research provides some evidence of the cognitive and learning benefits of employing multiple pedagogical agents, each assigned to distinct knowledge bases, in a multimedia learning environment. However, follow-up studies and extensions of these findings remain scarce. To address this gap, we draw on multimedia learning and cognitive models to investigate the effects of using multiple AI voices as specialist virtual tutors for distinct programming algorithm subtopics on cognitive load and learning outcomes. A between-subjects experimental design was employed with first-year business undergraduates who had minimal programming knowledge. Participants engaged with a multimedia learning video, narrated either by a single AI voice or by three distinct AI voices, each assigned to a different subtopic. Cognitive load was measured via a survey, while learning outcomes were assessed using immediate and 2-week delayed posttests covering retention, near-transfer, and far-transfer tasks. Results indicated that participants in the multiple AI voice condition reported significantly lower intrinsic and extraneous cognitive load compared to those in the single AI voice condition. Furthermore, the multiple AI voice group outperformed the single AI voice group in both immediate and delayed retention, as well as in immediate far-transfer tasks and delayed near-transfer. This study empirically extends prior research on the cognitive effects of using multiple AI voices as virtual tutors in multimedia learning environments. It offers preliminary evidence that using unique voices to distinguish subtopics can benefit cognitive load and learning outcomes, with theoretical and instructional design implications for leveraging AI text-to-speech engines to simulate multiple virtual tutors for distinct instructional topics.