Automated Gender Recognition in Visual Art

In this talk we will discuss the interplay between art and gender by integrating theoretical and methodological advances from the social sciences and artificial intelligence. We will begin the talk with an introduction to Computer Vision, the subfield of AI dedicated to visual analysis, to construct a common frame of reference and some familiarity with the techniques used.

We will then present our empirical work at the intersection of gender and art. Gender norms have undergone significant changes over the last century as a result of influential social movements. Importantly, the construct of gender has moved from a binary framework (male/man, female/woman) to a wide diversity of expressions that defy the gender binary and related gender stereotypes. At the same time, human figures have always featured prominently in visual artworks, providing a historical lens on how gender has been represented. Our research examines whether artworks influence the way we think about gender, or whether they merely reflect common conceptions of gender prevalent in a given era and culture.

While the scale of the data implies automation, the available automatic techniques for gender often still assume a binary framework, limiting their applicability. Therefore, we explore how we can perform analyses of visual art with such techniques, how they can be modified to go beyond the binary, and how this relates to human perception of gender and aesthetic appreciation of gender non-conforming artworks. In the talk we will present initial results of our project and discuss implications and future directions.