Research on 3D Modeling of Digital Textbooks Driven by Generative Design——Framework Construction and Effect Verification Oriented to “Cognitive Load Reduction”

Generative design boasts great potential in the 3D modeling of digital textbooks by lowering technical barriers, enabling precise resource provision, and realizing cognitive adaptation. However, due to insufficient understanding of instructional needs, the existing generative design of 3D models tends to fall into the pitfall of over-realistic simulation, which imposes excessive cognitive load on learners. Therefore, its design philosophy urgently needs to shift toward simplification..In view of this, based on five theoretical foundations including system science and five design principles such as hierarchy, this paper highlighted the pivotal role of prompts and constructed a four-stage generative design framework for 3D models, covering prototype definition, evolutionary generation, evaluation and selection, and iterative optimization, to achieve the goal of cognitive load reduction. Afterwards, an experiment was conducted with a Latin square design. Through descriptive statistics and multivariate analysis of variance, this paper compared four modeling modes including human-AI collaborative design with simplification, human-AI collaborative design with simulation, manual construction with simplification, and manual construction with simulation, in terms of learners’ cognitive load and learning performance. The results showed that the group adopting human–AI collaborative design with simplification under the proposed generative design framework achieved the best overall in reducing cognitive load, confirming the effectiveness of the framework. This study also underscored the prominent merits of simplified design philosophy for 3D models and human-AI collaborative modeling, offering methodological references for frontline teachers to utilize AI to generate learning resources in a scientific and efficient manner.