The micro-foundations of email communication networks

The popular and scientific literature has been discussing the advent of ‘big data’ with a measure of excitement and apprehension. For the first time in history, it seems, every breath we take, every move we make, someone’s watching us. But beyond their unprecedented volumes and the anxieties they raise, new communication data have a less obvious aspect, in so far as they are (arguably) of a fundamentally different kind, compared to traditional network datasets. Traditionally, social network data describe relationships between individuals; quasistatic social ties such as friendship, trust, kinship and employment relations. But when they are used to model digitally mediated communicative transactions, the connections are of a different nature. Instead of representing stable social ties, transactions (such as emails, text messages and phone calls) constitute sequences of shortlived events, with each transaction being a possible response to a preceding one and a potential stimulus to the next. The point of departure of this dissertation is the distinction between the topology of the tie structure and the temporal structure of sequences of communicative transactions. Theoretically, the dissertation explores mechanisms of co-evolution between these two structures at three levels of aggregation: (i) the macro-level consisting of the network itself or substructures within it, the level of an organization or a community as a whole; (ii) the meso-level consisting of nodes and social ties; and (iii) the micro-level consisting of sequences of interrelated communicative transactions. On the one hand, networks, individuals and ties are seen as the backdrop against which sequences of transactions unfold. On the other hand, transactions are considered to have (cumulative) consequences on the evolving structure of social ties and the network at large. Methodologically, the thesis uses a publicly available dataset consisting of email transactions within Enron, an American energy and services company, during the few months of its bankruptcy. Two methods are applied to identify and explore the mechanisms. First, the dataset is disaggregated into various types of email transactions, revealing how different transactions contribute to various structural properties of the network. Second, a multilevel analysis approach is used to reveal how structural and transactional mechanisms combine to elicit new communicative transactions on the part of email recipients. The mechanisms identified in the empirical chapters challenge received wisdom about the nature of social networks and their link to the notion of social (trans)action while at the same time addressing practical problems faced by network modellers who need to construct networks out of digitally mediated transaction datasets. In addition, the findings raise general questions about new types of data and the consequences they may have, not only for the field of social networks, but also for popular ways of thinking about ‘the social’ and ways of intervening in its course.