Gender-differentiated effects of motion-enhanced AI-generated images in instructional videos: a mixed-process study using TAM-related perceptions and cognitive load theory

With the widespread adoption of online educational technologies, instructional videos have become a pivotal tool for flexible and efficient learning. The integration of visual aids, such as animations and dynamic elements, aims to enhance the learning experience; however, their effectiveness remains debated. This study examined learning performance and subjective experience when instructional videos presented static AI-generated images or manually animated versions of those images. A 2 × 2 between-subjects experiment was conducted with 40 undergraduate participants ( n = 10 per cell), comparing static versus motion-enhanced presentation conditions across male and female participant groups. Stable Diffusion XL generated the base images, and CapCut was used to add motion; the motion itself was not AI-generated. Participants viewed a 140-s mobile video on mortise-and-tenon structures, completed a 10-item accuracy test, and rated willingness to continue learning, preference, acceptability, ease of understanding, and mental effort. The ANOVA results provided no statistically significant evidence of a main effect of motion on learning accuracy. Differences across gender groups were observed for several subjective outcomes. For acceptability, the gender × motion interaction was statistically significant, but Bonferroni-adjusted simple-effect tests supported a gender-group difference only in the static condition. For mental effort, the interaction was also statistically significant; adjusted simple-effect tests indicated higher mental effort among female than male participants in the dynamic condition and higher mental effort in the static than dynamic condition among male participants. Given the small sample and multiple outcomes, these findings should be interpreted as exploratory.