Moods, bots, and bodies: University students’ emotional and physiological responses to human vs. GenAI chatbots

Abstract

As generative AI (GenAI) chatbots become more common as learning partners, questions remain about students’ emotional and physiological responses to them. A multimodal design was used to compare university students’ experiences during a 25‑minute brainstorming session with either a human teacher or GenAI chatbot. 30 participants wore EmbracePlus sensors to record heart rate, electrodermal activity, and skin temperature while completing the task, and completed mood questionnaires before and after brainstorming. Analyses compared mood change scores and examined physiological data for both temporal patterns and total activation (area-under-the-curve; AUC). Although both groups reported increases in several positive moods, outcome-specific differences emerged, whereby students brainstorming with a human teacher showed greater gains in positive mood, whereas the chatbot group reported increased stress and discouragement. Although physiological change trajectories did not differ by condition, condition differences emerged for selected physiological indicators, and specific AUC measures were associated with mood outcomes: higher pulse AUC was linked to negative moods, and higher skin temperature AUC to positive moods. These findings provide preliminary evidence that human facilitation may produce stronger positive emotional outcomes, while GenAI chatbots can support meaningful physiological engagement and may serve as valuable complementary learning tools. Physiological signals also revealed associations between bodily states and emotional experiences, underscoring the value of integrating multimodal data into research on AI‑mediated education.