Cognitive Alignment Through Explainable AI: Supporting Designers’ Understanding of Typicality and Novelty in User Aesthetic Perception

This study presents a human-centered AI (Artificial Intelligence) system designed to support designers in interpreting how non-expert users cognitively perceive product form characteristics. Grounded in the Unified Model of Aesthetics (UMA), the system classifies personal computer (PC) designs along the cognitive dimensions of typicality, novelty, and MAYA balance, using an improved YOLOv11s framework. It outputs interpretable feedback through Grad-CAM visualizations and textual rationales to enhance designer understanding and support perceptually aligned design decisions. In Phase I, the model showed high alignment with non-designers rated novelty and moderate alignment with typicality, supporting its cognitive validity in capturing perceptual salience and identifying schema-based form expectations. In Phase II, designers using the system achieved higher Top-2 cognitive alignment and Spearman correlations, reporting greater confidence, usability, and empathy toward non-designers. Thematic analysis of post-task interviews confirmed the system’s transparency and its role in bridging expert and non-expert perception gaps. By integrating explainable AI with cognitive aesthetics, this research contributes a replicable framework for perceptual mediation in early-stage design. The system enhances human–AI collaboration by facilitating interpretability, fostering empathic alignment, and promoting inclusive design reasoning.