Semantic Based Clusters of VK Users Avatars and Their Association with the Big Five Personality Profiles

Online social media have become one of the key sources of data about one’s personality due to its availability and popularity. One of the major research questions is how the user personality reveals itself in digital traces. The avatar images and its graphical features are at the focus of the study. The CLIP image encoder was used for feature extraction, that were used for clusterization of VK users avatars. The research hypothesis that VK users with different semantic content of their avatars have different Big Five scores was verified. The data were gathered from the VK online social media with a specially developed app.