The relationships among facial features and impressions: statistical causal discovery using LiNGAM

Numerous studies have explored the factors that shape facial impressions. However, the details of the relationships among various facial features, such as facial morphology, skin, and impressions, remain vague. This study aimed to explore the relationships among facial morphological features, image features, attractiveness, sexual dimorphism, dominance, and trustworthiness. We applied the Linear Non-Gaussian Acyclic Model (LiNGAM), which statistically extracts hypotheses of causal relationships among variables in a data-driven manner. The results confirmed that morphological features, such as facial density, and image features, such as fractal dimension and spectral slope, affect facial impressions. Additionally, we found a causal path from each impression to trustworthiness, and the path for the mediation of attractiveness and sexual dimorphism differed by image sex. The results suggest that this hypothesis-independent exploratory method may clarify the relationships among facial features and impressions. Ultimately, the approach contributes to the development of a psychological model and advances understandings of the links between facial features and impressions.