Essays on Social Media Content Delivery and Engagement

Abstract

This dissertation studies how social media content delivery across channels and formats affects user engagement. The first chapter investigates the effects of transience on content creation and consumption and how social media influencers can use permanent and transient channels to maximize follower engagement. Using computer vision and deep learning methods, we extract image quality and embeddings from a novel dataset of Instagram content to compare transient and permanent content and investigate how such differences affect follower engagement. The findings reveal that influencers can increase engagement by publishing more professional content in permanent channels and more personal and authentic content in transient channels. The second chapter studies how consumers perceive time when watching long-form videos vs. short-form videos. We find that viewers who successively watch a series of short-form video clips underestimate total time spent watching videos, compared to viewers who viewed the same video content and length presented as one long-form video. This time underestimation effect occurs because shorter videos naturally induce more viewer action to continue video consumption, i.e., clicking or swiping to the next video. These content-unrelated actions divert viewer attention from the actual video content, leading to lower perceptions of time spent on watching videos. Together, the findings demonstrate one explanation for the increased consumption of short-form videos in recent years: when watching short-form videos, consumers are more prone to losing track of time and spend more time than they anticipate. The final chapter concludes with an assessment of social media moving forward, specifically on the implications of GAI on influencer marketing. Incorporating interviews with an influencer and a digital talent agency, I describe AI use cases in influencer marketing as well as benefits and challenges that it poses for influencer marketing. The chapter concludes with suggested directions for future research.