丝绸2025，Vol.62Issue(5)：1-12,12.DOI:10.3969/j.issn.1001-7003.2025.05.001
人工智能生成式纹样的审美评价研究
A study on the aesthetic evaluation of AI-generated traditional ornamental patterns
摘要
Abstract
Traditional patterns are important cultural carriers and aesthetic symbols of the Chinese nation,with profound significance in conveying historical narratives and national identity.Artificial intelligence generation technology provides a new path for innovation in traditional patterns.However,current research on these intelligent generative patterns is mostly focused on innovation and optimization at the level of generation technology,and there is still insufficient research on their aesthetic aspects.There is a lack of quantitative evaluation systems for intelligent generative patterns from aesthetic perspectives.This study focuses on the Xilan Kapu patterns of the Tujia ethnic group,comparing the pattern works generated by AI(Human-Out-of-the-Loop,HOTL),the pattern works jointly designed by AI and humans(Human-In-the-Loop,HITL),and the ethnic traditional patterns created by humans.Through statistical methods,participants' different responses on the three types of patterns in terms of aesthetic perception and cognitive discrimination were analyzed to explore the aesthetic evaluation of artificial intelligence art. The experimental results demonstrate that in terms of aesthetic perception,traditional ethnic patterns and human-intervened HITL generated works achieved comparable evaluation scores,both significantly outperforming HOTL outputs.Through the integration of linear mixed-effects modeling and SHAP interpretability analysis,we quantitatively identified 11 statistically significant determinants influencing the perceived"beauty"of patterns,thereby establishing a dedicated aesthetic evaluation framework for Xilan Kapu generative motifs. Regarding cognitive attribution,participants predominantly assessed traditional ethnic patterns through cultural-connotative dimensions when judging their authorship,while evaluating AI-generated patterns based on regularity-based criteria.Notably,a negative correlation emerged between attribution accuracy and aesthetic ratings for AI-generated patterns.Despite participants' inability to accurately identify the creators,they consistently attributed patterns with higher aesthetic scores to human authorship,revealing an unconscious anthropocentric attribution bias in aesthetic valuation. This study demonstrates the potential for collaboration between humans and AI in creation.Through human-machine cooperation,computers can unleash their powerful generation capabilities while optimizing the quality of works based on human aesthetic judgment.It enriches the theoretical foundation of AI in the field of traditional pattern generation and aesthetic evaluation,and provides new ideas for enhancing the aesthetic attributes of generative patterns,enabling technology to truly serve the dynamic inheritance of traditional culture.关键词
人工智能生成/西兰卡普纹样/审美评价模型/认知判别/审美感知/神经美学Key words
artificial intelligence generation/Xilan Kapu patterns/aesthetic evaluation model/cognitive discrimination/aesthetic perception/neuroaesthetics分类
轻工业引用本文复制引用
孙家豪,程子悦,胡锦霞,朱诗源..人工智能生成式纹样的审美评价研究[J].丝绸,2025,62(5):1-12,12.基金项目
湖南省自然科学基金项目(2025JJ60816) （2025JJ60816）
长沙理工大学研究生科研创新项目(CLKYCX24093) （CLKYCX24093）
湖南省教育厅重点项目(22A0203) （22A0203）
教育部产学合作协同育人项目(231107610085945) （231107610085945）
