Notes
Digital Emotion Contagion in Online Environments: An Automated Content Analysis of Comments from Self-Transcendent YouTube Videos
138Self-transcendent emotions (STEs) such as admiration, hope, or gratitude have many beneficial effects on media users and can be elicited by media content (Oliver et al., 2021)—including online videos (e.g., Dale et al., 2017). Digital emotion contagion literature suggests that viewers’ emotional responses to online content can converge with those expressed by others (e.g., in video comments) (Goldenberg & Gross, 2020), potentially creating a positive, reinforcing environment. While user comments were indeed found to affect viewers’ responses to STE-inducing videos (Krämer et al., 2021), research on the comments themselves is sparse (Oliver et al., 2021). As STEs stimulate a need to share the experience (Ji et al., 2019), it seems likely that users would report video-induced STE responses in the comments. Hence, we expect a high congruence between specific STEs expressed in the video and those expressed in the comments (H1). Moreover, viewers may like or reply to comments, which increases comments’ visibility and potential impact. As viewers identify mostly with commentators that express emotions in line with STE-inducing videos (Krämer et al., 2021), we expect congruent comments to receive more engagement (H2).
139This study examines 53,433 English language top-level comments from 86 videos coded for STE elicitors by Dale et al. (2017). We automatically coded STEs expressed in these comments using the STE-dictionary (Ji & Raney, 2020) and EmoRoBERTa (Ghostal, 2021)—a transformer model (Vaswani et al., 2017) trained on the GoEmotions dataset (Demszky et al., 2020). Our gold standard validation (n = 200 comments coded by two coders) demonstrated better reliability for the transformer (Agreement .82–.95) than the dictionary (Agreement .71–.92).
140In our analysis, we will correlate the proportion of scenes depicting an elicitor (coded by Dale et al. [2017]) with the proportion of comments expressing that emotion (H1) and test the influence of content-comment congruence on engagement (H2).
141Self-transcendent emotions / emotion contagion / online videos / user comments / computational methods