Reviewing the Emerging 'Big Social Data' Research

The next session at Social Media and Society is on 'big data', and begins with Andra Siibak (who is also the programme chair for AoIR 2017 in Tartu, Estonia!). She highlights the possible methodological shifts that arise from the use of 'big data' in social science research: this is in part seen as a shift towards more quantitative methods, but also as a more nuanced and methodological shift from designed to more 'organic' data, whatever we may mean by this. Approaches that are built on formulating and testing preconceived hypotheses may also be challenged by other, alternative approaches.

Andra and her colleagues carried out a systematic literature review to understand what kinds of methodologies are actually being used in social science studies drawing on 'big data'. The majority of such studies were from media and communication as well as computer science (but also involving significant interdisciplinary collaboration), and there has been a tenfold growth in articles from 2012 to 2015. The vast majority worked with Twitter data, and drew on unstructured or semi-structured datasets that needed to be further processed ahead of analysis. Data fusion between several different datasets was uncommon.

Some 55% of articles still commenced with a preset hypothesis, while 32% examined pre-conceived research questions; the problem setting was also usually very strongly tied to existing theory – so the 'end of theory' that had been predicted in the context of 'big data' has not yet arrived. However, the vast majority of studies were exploratory and/or predictive, too.

Key methods being used included statistical, content analysis, computational, and social media analytics techniques. There is therefore a strong focus on quantitative approaches, it appears. Authors often claimed novelty for their research methods, tools, technologies, and techniques, but also for their theoretical frameworks. Andra suggests that such data-driven theory-building activities will further increase over time, and that the combination of emergent computational and traditional manual analysis methods will also become more prevalent. It will also be useful to explore knowledge transfer processes by analysing the theoretical and methodological groundwork that these articles reference.