Exploring Cultural and Gendered Self-Presentation in Online Dating through Image Clustering

Research on user behaviour and motivations in online dating is prevalent across disciplines, but the actual self-presentation of users in practice is frequently overlooked. An in-depth analysis of self-presentation and associated practices in picture-based online dating will provide a key understanding of online dating’s self-regulatory potential and the dominant self-presentations that are regulated by it. The main research question of this analysis is how online dating users present themselves through configurations of profile pictures and whether there are cultural and gendered differences. This study focuses on applying unsupervised, machine learning-based image clustering to identify visual clusters of self-presentations in picture-based online dating. In doing so, this study will explicate and describe the visual discursive space of picture-based online dating, as well as the variations across culture, gender, and gender affinity for social interaction. This paper first introduces the study’s data format and unique challenges in clustering online dating images and profiles. It then reviews the workflow of the employed machine learning-based image clustering. The major discussed aspects are (1) feature mapping, (2) dimension reduction, (3) vector pooling, and (4) the clustering itself. Subsequently, the findings of the clustering based 13 countries across 5 continents, as well as gender and target audience orientation will be introduced. The results of the illustrated image clustering will allow for an understanding of the culture and gender-specific nuances in self-presentation. These results will also serve as a basis for further qualitative analysis in successive studies.