Digital traces, sociology and Twitter: between false promises and real potential

We are told that society changes. It evolves toward a more fluid, active and horizontal form of socialisation (Bauman, 2000; Castells, 2009; Sheller & Urry, 2006; Urry, 2000; Wittel, 2001). As much as a change in the social, it is also a change in the conception sociology gives to the sociality and society as large. There is a shift in social theory with recent raises of post-demographic perspective, and postmodernism (Latour, 2005, 2011, 2013; Ruppert, Law, & Savage, 2013).Along these, post changes, another revolution takes place, in the use of new technology and mainly the Web. These new uses are associated with an explosion of the quantity of digital traces, often labelled as Big Data.This Big Data paradigm brings radical changes in research that are perfectly suitable for computational researchers and other data scientists, and they are taking full advantage of it (Hey, Tansley, & Tolle, 2009). This new landscape, in society and in research, puts pressure on sociology, and other social sciences fields, to find adequate answers to these societal, theoretical and methodological challenges. The best proxy of these challenges and the tensions between scientific fields and the new form of social interactions are the Social Network Sites (SNSs). They represent the extreme case of horizontal and fluid interactions while producing an incredible amount of accessible digital traces.These digital traces are the essential bricks for all the research using SNSs. But despite this importance, few researches actually investigate and explain how these digital traces are produced and what is the impact of their context of access, collection and aggregation. This thesis focuses on these digital traces and the gap left in the literature. This empty space is however conceived as a central point of tension between sociological positions on how to define new social interactions, and methodological principles imposed by the logic of Big Data.The work is articulated around one specific social network, Twitter. The reason for this choice lays in its openness, the easy use of its APIs, and, in consequence, by the fact it is the most extensively studied SNS for now.I begin the work on the definition of the new form of sociality using the network concept as the key concept around which several notions, such as social, cultural and technological can be articulated. I conclude that none of these evolutions are independent and need to be seen as co-integrated. In consequence, the change in the social interaction needs to be seen as much as a factual change, than a change in our way to interpret it. From this conception of the network and the importance in our understanding of social interactions, I retrace the evolution of the notion of Big Data, specifically with the example of Tesco and their ClubCard. This is the first step to locate the technological changes into a more comprehensive methodological framework. This framework, the transactional perspective, is decomposed to understand the consequence of such position applied on SNSs and specifically on Twitter. This is the first explanation of why the research almost entirely focuses on the Tweets and what are the consequences on our understanding of the interaction on the social web service. Then I use this first iteration of the definition of a digital trace to build a new definition of what a Social Network Site is and centre this definition around the concept of activity and context.I operationalise these concepts on Twitter to develop a new method to capture social interaction and digital traces that are often put aside due to the difficulty of their access. This method takes into account the limits imposed by the Twitter APIs and describes the consequences they have on the generation of a dataset.The method is based on a constant screening of sampled profiles over time. This method allows people/us/you to reconstruct the missing information in the profile (the trace of the changes in the friends’ and followers’ lists). This information creates the measure of context (the user’s network) and activity (tweeting and adding or removing links) defined earlier.The obtained dataset provides an opportunity to see the importance of the aggregation process and the flexibility offered by the digital traces. Then following this, I developed three analyses with different levels of aggregation for different purposes. The first analysis was to test the hypothesis of the influence on users’ context activity on their own activity over time. The second analysis, did not use the time as a measure of aggregation but tested the same hypothesis on an individual level. And finally, the information about the activity itself is analysed in order to see to which extent the digital traces obtainable contain the sufficient information about the change in activity itself