Why people press "like" : A new measure for aesthetic appeal derived from Instagram data

Abstract

Be it on Facebook, Twitter, or Instagram, “Like” buttons are all over social media generating huge amounts of data. In this project, we develop methods for leveraging Instagram data with the purpose of developing a measure that is useful as a proxy for the aesthetic appeal of photographs. Based on the metadata of 15,073 photographs from the photographic genres of architecture, dance, and landscape gathered from 9 different Instagram accounts of professional photographers, we compute the Image Aesthetic Appeal score (IAA). We conduct an online experiment to test how IAA scores relate to more commonly used psychological variables, such as rating scales of aesthetic liking. We also investigate both low-level features and content-related preferences in the image set. Our results show that IAA scores are a reliable measure and are substantially predictive for experimental liking data. Also, solidly established preferences for curvature and visual balance as well as content-related preferences are reflected in IAA scores. We provide a simple method to compute IAA scores for all sorts of content on Instagram. This prepares ground for investigating large data sets with regard to aesthetic appeal of photographic images.