Human–AI Collaborative Decision-Making for Ceramic Cultural Heritage Product Design: An Integrated AHP–AlexNet Framework

Abstract

Against the backdrop of globalisation and accelerating digital transformation, the creative redesign of ceramic intangible cultural heritage (ICH) encounters persistent challenges, including inefficient subjective evaluation and the difficulty of simultaneously preserving cultural significance and achieving visual appeal. To overcome these limitations, this study develops an integrated decision-support framework that combines the Analytic Hierarchy Process (AHP) with the AlexNet convolutional neural network (CNN) to improve the evaluation and selection of ceramic ICH creative products. The research first establishes a multidimensional assessment framework encompassing cultural inheritance, visual aesthetics, human-centred ergonomics, and innovation potential. By integrating Kansei Engineering with AHP, the framework ensures that design decisions remain aligned with traditional cultural values while satisfying engineering ethics and functional requirements. Subsequently, a transfer learning approach is employed to optimise the AlexNet model using a purpose-built ceramic aesthetics dataset comprising 24,000 high-quality samples. This enables the deep learning model to perform rapid and objective quantification of complex visual characteristics. A comprehensive comparative evaluation is conducted against widely adopted deep learning architectures, including VGG-16 and ResNet-50. The findings demonstrate that AlexNet delivers the most favourable trade-off between computational efficiency and aesthetic feature extraction accuracy, achieving a validation mean squared error of 0.4912 and a Pearson correlation coefficient of 0.875. Furthermore, the ablation experiments reveal that the proposed subjective-objective score mapping mechanism effectively mitigates both the cultural disconnect associated with purely algorithmic evaluation and the subjective inconsistencies inherent in conventional manual assessment during the selection of 100 conceptual blue-and-white porcelain stationery designs.  Overall, the proposed framework establishes a comprehensive human–machine collaborative decision-making approach that spans theoretical development, implementation methodology, and quantitative performance evaluation. The study offers a robust and scientifically grounded solution for ceramic craftsmanship, creative product design, and the wider digital transformation of ICH.