RETHINKING ENGAGEMENT IN AI-DRIVEN ONLINE LEARNING THROUGH HUMAN EMOTION

Artificial intelligence is now embedded in many online learning systems. These systems adapt, decide, and give constant feedback. The key question is how to model learner engagement in such contexts. Conventional studies rely on flat models that assume functional usability or convenience translate directly into engagement. This study contends that such an approach is insufficient and examines an alternative based on the Stimulus-Organism-Response (SOR) framework. External stimuli from the system affect internal states, which then drive behavior. Using PLS-SEM with AI-experienced learners, it contrasts a direct-path model with a mediated model in which motivation and fatigue act as internal mechanisms. Findings show that while flat model approach captures variance, it obscures process. The SOR approach reveals that performance expectancy and self-efficacy enhance motivation, whereas learning anxiety and feedback overload intensify fatigue. Motivation drives engagement; fatigue diminishes it. Significant indirect effects confirm that engagement in AI-mediated learning is psychological before it is behavioral. The study concludes that effective system design requires environments that are not only intelligent but also emotionally and cognitively responsive.