Temporal dynamics of students' physiological arousal in personalized learning contexts: a quasi-experimental study based on electrodermal activity.

Physiological arousal varies throughout classroom learning, yet its temporal dynamics during personalized learning in authentic primary-school classrooms remain unclear. This quasi-experimental study used electrodermal activity (EDA) to examine the lesson-wide temporal trend, dynamic trajectory, and phase-specific arousal level and activation frequency in an "Introduction to Circles" lesson. Two intact fourth-grade classes in Shanghai participated (analytic sample: experimental n = 23; control n = 27). The experimental class received personalized learning based on cognitive diagnosis, targeting discrepancies between students' existing conceptions and target concepts; the control class received conventional instruction. Raw skin conductance (SC) and skin conductance response frequency (SCRF) served as operational indicators; within-person min-max standardized SC (SSC) served as a robustness check. Mann-Kendall tests, individual-level linear mixed-effects models, and matched-time comparisons were conducted. Across 17 two-minute windows, SSC declined significantly in both classes (control: Z = -2.266, p = 0.023; experimental: Z = -4.490, p < 0.001). The raw-SC model showed a significant Condition × Time interaction at the two-minute resolution (p = 0.008), but not at four or six minutes. Of four event-boundary adjacent-window comparisons, two Condition × Phase interactions remained significant after Holm correction: SC increased in the experimental class and decreased in the control class at the first episode start (adjusted p = 0.012); the pattern reversed at the second episode end (adjusted p = 0.002); both were reproduced using SSC. Phase-average raw SC did not differ between conditions (p = 0.900), whereas SCRF was higher in the experimental class (p < 0.001, d = 2.76). Post-test mathematics achievement was also higher in the experimental class (p = 0.005). Physiological arousal declined as the lesson progressed; the personalized-learning condition did not consistently alter this trajectory but the two classes showed different short-term SC changes at two instructional transitions. The higher post-test achievement suggests a potential learning benefit; however, the relationship between these physiological patterns and the achievement difference warrants further investigation.