Contributors:
- Guillaume Latzko-Toth (panel chair), Assistant Professor, Department of information and communication, Université Laval
- Claudine Bonneau, Assistant Professor, Department of management and technology, Université du Québec à Montréal (ESG-UQAM)
- Mélanie Millette, Substitute Professor, Department of social and public communication, Université du Québec à Montréal
In the context of a “computational turn” in social sciences and humanities (Berry, 2011), and a growing appeal for “Big Data” analysis in those fields, is qualitative research based on small samples and corpuses still relevant? Traditional ethnographic approaches are deemed to provide insights based on a limited number of observations or, worse, declarative accounts of practices relying on participants’ memory—and honesty. “Big Data” methods, on the other hand, promise to grant researchers “direct” access to real practices of vast populations. However, as noted by some critiques (boyd & Crawford, 2012; Gitelman, 2013; Tufekci, 2014), interpreting and making sense of these loads of data can be very challenging, especially considering the importance of context in the study of media practices. Others have pointed out that rather than opposing them, Big Data and “Thick Data” strategies can work together (Wang, 2013; Ford, 2014). In the three presentations of this panel, social scientists with a qualitative, interpretive approach share some innovative strategies of inquiry they devised for case studies on the uses of social media.
In his presentation “Small data, thick data: (Re)collecting networked publics’ experiences of a public debate,” Guillaume Latzko-Toth presents an innovative method that Madeleine Pastinelli (Université Laval), Nicole Gallant (Institut national de la recherche scientifique, INRS) and he have crafted for their research on the circulation of information on social media during the 2012 student strike in Quebec. They wanted to know to what extent Facebook had been used by young adults to be informed, discuss, and form their opinions on this major public issue. After having considered the option of using a computer script to collect whole sets of digital traces from a sample of Facebook users, an alternative approach was developed: the “commented visit” of users’ activity logs. A series of semi-structured interviews were conducted with young adults (students and non-students, N=30). While the first part of the interview was conventional, with a series of topics ranging from informational habits to political engagement, the second part took place in front of the computer screen, and consisted in the examination of the participant’s Facebook activity log around predetermined dates, over a whole year. The posts, comments, “likes” and other forms of interactions with contents and other users were video-recorded with a dynamic screen-capture software and analyzed along with the participant’s oral comments during the visit. This method allowed researchers to observe digital practices retrospectively, with the benefit of contextualization and reflexivity from the subjects themselves, adding a layer of “thickness” to the data.
In her presentation “Investigating open ended phenomena on Twitter through manual data collection: Challenges and opportunities,” Claudine Bonneau argues that Big Data approaches for collecting user-generated content are of no use when machine-processable criteria cannot be predetermined. This is the case with an online phenomenon referred to as “working out loud”, an emergent practice that may be described as a process of continuously narrating the work during the course of its realization. Because that practice can emerge on Twitter as employee-driven initiatives and can be found in any professional field, it is not possible to a priori circumscribe data collection to specific organizations where it is encouraged or prescribed by the employer, nor is it possible to a priori limit investigation to one professional community where working out loud would already be quite common. Such an open-endedness implies that it is impossible to specify any events, hashtags or keywords. While posing a series of methodological challenges, the manual collection and qualitative analysis of a modest corpus of tweets (N=200) allowed Bonneau to generate research intuitions that could not be discovered through a quantitative analysis of a large dataset, namely the role that Twitter can play in workers’ lives.
In her presentation “From small data to thick data: Investigating Twitter uses in French-Canadian minority context,” Mélanie Millette explains the challenges she faced while she was investigating how Francophones outside of Quebec were using Twitter to become visible. Although Canada’s French-speaking minority is largely concentrated in Quebec, over a million French-Canadians live outside this province. This minority has developed social media strategies to gain visibility in the mediated public sphere. Considering the scattered nature of the population under study, a mixed method and a long time span were key to collect relevant, “thick” data. Although the methodology was partly inspired by data mining techniques that often result in “big” datasets, Millette collected a relatively “small” set of tweets over four months (N=8,764). Three main strategies of inquiry have been mobilized. First, observant participation on Twitter was conducted for more than 2 years and a half. Second, a mixed method was crafted where data mining was guided by a first round of exploratory interviews (by phone) and by the observation phase. Third, a second set of interviews has been conducted in person (N=27) with users selected from the Twitter database. Ultimately, the method succeeded in providing thickness and depth to the analysis, as data from Twitter mining and data from interviews were informing each other in terms of social media patterns and meaning in the French-Canadian minority context.
