The influence mechanism of AIGC-driven “learning-thinking-applying-creating” integrated animation instructional design on college students’ cognitive load and creative thinking

Abstract Generative artificial intelligence has entered animation classrooms faster than research has been able to describe what it does to learning. This study developed a “Learning-Thinking-Applying-Creating” (LTAC) integrated teaching model built on an open-source AIGC toolchain and evaluated it in a twelve-week quasi-experiment with second-year animation students, two intact classes being allocated to an experimental or a control condition. Cognitive load was measured with a three-subscale instrument adapted from Paas and from Leppink and colleagues, and divergent thinking with an abbreviated Torrance-type battery scored for fluency, flexibility, originality, and elaboration by two trained raters who worked blind to group and occasion (intraclass correlations 0.86 to 0.91). Data were collected at pretest, mid-test, and posttest. Analysis of covariance indicated that the intervention lowered extraneous load and raised germane load while leaving intrinsic load unchanged. Repeated-measures analysis revealed selective gains: originality and elaboration improved more in the experimental group, whereas fluency and flexibility rose in both. Structural equation modelling was consistent with a differentiated mechanism, in which lower extraneous load predicted fluency and flexibility, germane load related to originality and elaboration indirectly through a deep-processing strategy, and intrinsic load showed an inverted-U association with the creative-thinking composite. Because the outcome measure is domain-general, the results speak to divergent thinking rather than to animation ability as such. The study extends cognitive load theory into generative-AI-supported creative education and offers animation programmes a reproducible instructional design that lowers production barriers while protecting deep creative work.