Collaboration Between Social Sciences and Computer Science: Toward a Cross-Disciplinary Methodology for Studying Big Social Data from Online Communities

Abstract

Online communities are now extremely numerous. Most of them being multifaceted, dynamic, and rapidly evolving, they are of the utmost interest for social science researchers. One of the special characteristics of these communities is the production of numerical traces generated by communications between members, inside their communities or through social networks. These traces, captured and stored by software managing their dissemination, represent a massive amount of data. Based on their volume, velocity, variety, and veracity, they must be handled in the context of the big data phenomenon. These novel constraints generate scientific, epistemological, and ethical problems related to the limited understanding researchers have of the algorithms utilized by software tools, their possibilities and limitations, error rates, and biases. As a consequence, social science researchers interested in mining all these data often depend on data analysts who lack any social science background. Collaboration between social sciences and computer science is hence critical to meet these challenges, and to propose a cross-disciplinary methodology combining the contributions of both fields towards the study of online communities. Using online communities of video game players as an example, this contribution puts the emphasis on identifying the challenges associated with the study of online communities, and proposes a methodology combining computer science and social science approaches. First, we present research questions, categorizations, and classifications related to identity, communication, and social dynamics by linking them to data mining and automated processing techniques. We then study how to integrate social science models into computer tools, and link qualitative methods with big data analysis in order to overcome errors in the interpretation of results related to data decontextualization. Finally, we formalize ethical concerns of social science researchers regarding limitations of software tools. This chapter hence demonstrates the scientific, epistemological, and ethical advantages of combining accepted methods from computer science and social sciences in order to propose a cross-disciplinary methodology for research on online communities.