Examining Flow Through Facial Expression-Based AI: Effects of Learning Styles and Instructional Contexts

This study investigates how learners’ cognitive styles and instructional environments influence flow, using facial expression-based AI to detect engagement-related emotional states. Based on Kolb’s experiential learning theory, participants were categorized into four styles: assimilating, converging, accommodating, and diverging. Instructional contexts were defined by format (individual, group-based, blended format) and delivery mode (offline, online, blended learning). Over 1 million repeated observations from 2,175 adult learners were analyzed using multinomial logistic regression to examine variations in flow states: focused, distracted, or drowsy. Results showed that learning styles significantly predicted flow. Accommodating and diverging learners were more likely to experience distraction or drowsiness than assimilating learners. Group-based formats supported flow more effectively for these learners, while individual formats were linked to higher disengagement. Additionally, accommodating learners were more prone to drowsiness in blended delivery environments, possibly due to interruptions in cognitive flow. These findings highlight the need for instructional designs that reflect individual cognitive preferences. This study also illustrates the utility of facial expression-based AI for large-scale measurement of learning flow.