Dimensions of Perceived Humanness in Artificial Faces: Evaluating Anatomical and Affective Impressions in the Uncanny Valley

This study investigates how humans evaluate artificial faces in the context of the uncanny valley, with the aim of exploring perceptual factors that may inform user-centered design. Participants viewed a series of facial images, including a real person, an AI-generated image, and several 3D-modeled faces based on the real individual. The Semantic Differential (SD) method was used to assess affective impressions, and a free-response question elicited participants’ intuitive criteria. Factor analysis revealed three interpretable dimensions: “Interpersonal Attraction,” “Impression,” and “Humanness.” Principal Component Analysis (PCA) further suggested that “Humanness” involves two subcomponents: perceived anatomical correctness and affective warmth. Text mining of the open-ended responses identified references to emotional expressions and facial realism as common bases for judgment. While the study remains exploratory, the results suggest that anatomical consistency and emotional cues may influence perceived familiarity and comfort in artificial faces. These findings provide tentative implications for affective computing and character design by highlighting aspects of facial presentation that may reduce negative impressions associated with the uncanny valley. Such insights may contribute to the refinement of user experiences in virtual agents and human-robot interaction, particularly when designing anthropomorphic characters in a human-centered framework.