Identifying Emotions in Images and Their Effects on Donation Behavior in Online Crowdsourcing Platforms

Abstract

Although textual emotional appeals in a donation context have been studied in past research, there has been little work looking at facial emotions expressed in posted images. Drawing on a panel data of 25,321 crowdfunding projects from Gofundme, we investigate how facial emotions expressed in posted images might be strategically used to increase individual project effectiveness as measured by metrics relevant to the platform, i.e., donation amount per project. In this research, we focus on four emotions that can be inferred accurately from facial expressions in images using artificial intelligence methods; happy, sad, anger and surprise. Our empirical analysis shows that increasing the degree to which facial expressions in an image convey any of these four emotions has a positive impact on the contributed amount to a project, with sadness having the biggest effect, followed by anger, surprise, and happiness. We also use MTurk studies to explore the process behind these effects, finding that perceived justice is the most dominant mediator in all the conditions of surprise, angry, happy and sad images, while perceived empathy and emotion contagion are mediators only in the case of surprise and happy images. And finally in a follow-up controlled study, we provide strategic recommendations for platform revenue management.