Markus Kaakinen, University of Tampere
David Garcia, Atte Oksanen & Pekka Räsänen
The pervasive use of the digital communication technologies has in many ways changed the patterns of social interaction in the Western world. Individuals can create and form social relationships online and build their own interest groups in ways that were previously not possible. Although researchers have studied online communities already during the early period of the internet, the rise of social media has rapidly transformed the dynamics of the online world, out-pacing research of its built-in effects.
The leading social network sites (SNS) attract millions of users and continue to expand simply by offering a platform where users can discuss, share and edit material of their own interests. For instance, it has been suggested that traditional peer groups have been joined by easily accessible computer-mediated groups, which have become an integral part of everyday life. Indeed, empirical results from comparative studies show that young people identify equally strongly with their online communities as they do with their own families, and even more strongly than with offline hobby groups. These communities are of particular importance for marginalized individuals, who face challenges to get social support and recognition from traditional offline environments.
Social interactions between users is delayed, yet capable to arouse strong collective emotions. Online users generally form strong social bonds with each other. The social interaction in online world is often regular and very normative. The norms and rules for discussion come predominately from the users. It thus follows that the conditions of interactions can change rapidly, for instance after some user posts an inappropriate or otherwise different type of comment. This emotional expressiveness has been found to both motivate users to participate and also to establish durable links to online communities. Both negativity and positive probably fuel online conversations, but they also may meditate the structures of discussion.
In our paper our interest lies in online interactions of eating disorder communities. Some of these communities focus on social support and health information sharing whereas so called Pro-eating disorder (Pro-ED) communities approach eating disorders as a lifestyle question, rather than as a medical condition. In these communities, it is typical for users to encourage eating disordered behaviour such as extreme dieting. Pro-ED phenomenon creates resistance in other online users. The forms of the resistance vary from publishing of counter material to stigmatizing and even dehumanizing debates.
We empirically study how positive and negative sentiments are expressed in the eating disorder discussions of the Reddit participatory news social network. Using the Reddit platform's interface, we automatically retrieved a data set of the full history of interaction in 25 eating disorder discussion areas (Subreddits), composing a data set with 4476 discussions and 28424 comments. On this data set, we apply social network analysis and sentiment analysis with the SentiStrength software, a state of the art unsupervised sentiment analysis technique that performs closely to human level accuracy for online comments and discussions.
Our results show that that the emotional expressions and the network structures of the discussions are significantly associated with each other. In general, eating disorder discussions with average measures of both positive and negative sentiment expressions generate most active commenting, while Subreddits and individual posts with extreme positive or negative sentiment measures generate less active commenting. Methodological possibilities and limitations of the SentiStrength analyses in the context of Reddit forum are also discussed.