Not a Mirror, but an Engine: Digital Methods for Contextual Analysis of “Social Big Data”

Digital media infrastructures give rise to texts that are socially interconnected in various forms of complex networks. These mediated phenomena can be analyzed through methods that trace relational data. Social network analysis (SNA) traces interconnections between social nodes, while natural language processing (NLP) traces intralinguistic properties of the text. These methods can be bracketed under the header “social big data.” Empirical and theoretical rigor begs a constructionist understanding of such data. Analysis is inherently perspective-bound; it is rarely a purely objective statistical exercise. Some kind of selection is always made, primarily out of practical necessity. Moreover, the agents observed (network participants producing the texts in question) all tend to make their own encodings, based on observational inferences, situated in the network topology. Recent developments in such methods have, for example, provided social scientific scholars with innovative means to address inconsistencies in comparative surveys in different languages, addressing issues of comparability and measurement equivalence. NLP provides novel, inductive ways of understanding word meanings as a function of their relational placement in syntagmatic and paradigmatic relations, thereby identifying biases in the relative meanings of words. Reflecting on current research projects, the chapter addresses key epistemological challenges in order to improve contextual understanding.