The Applicability of LiNGAM to Statistical Causal Discovery

Several studies have examined the factors that shape facial impressions; however, the relationships between facial morphology, skin, and facial impressions remain unclear. This study examined the relationships between facial morphology, image features, and impressions such as attractiveness, sexual dimorphism, dominance, and trustworthiness. Using a linear non-Gaussian acyclic model (LiNGAM), which is a data-driven method for causal inferences, we found that face density, fractal dimension, and spectral slope influenced facial impressions. The causal pathways to trustworthiness varied across male and female images and were mediated by attractiveness and sexual dimorphism. This exploratory method clarified the role of facial features in impressions, aiding the development of psychological models and the understanding of these relationships.