The People’s Internet Project and Its Struggle with Big Data

I’ve spent the morning in an AoIR Executive meeting, but I’m back for the second session on this Friday morning at AoIR 2018 – and I also have a paper in this session. First off is Rasmus Helles, though, who presents the People’s Internet Project: a major global study, supported by the Carlsberg Foundation, that seeks to map out global variations in Internet development.

This takes the citizen as a point of departure, and employs a range of methods for studying Internet use: it uses big data on Web traffic from ComScore; engages in local ethnographies of Internet users; conducts surveys of Internet and media use; and undertakes document analysis on Internet regulation and other forms of government activity in each of the countries it covers.

But this multi-layered study also creates a range of problems: the irony of big data is that everything is being traced and tracked, but these proprietary datasets are not consolidated, and almost nothing is immediately available to researchers. Further, survey questions are often vague, and therefore produce inaccurate answers, and similarly, ethnographies are always highly localised and difficult to compare one-for-one.

To address and mitigate these problems, the project has pursued a strategy of complementarity: the various datasets are able to complement each other, and at least point to high-level patterns of cultural and commercial interdependencies; the ComScore data, for instance, obviously only covers Web browsing patterns, and the surveys can provide further detail on the use of other media channels and devices beyond the Web itself.

A second strategy is corroboration: big data and ethnographies may corroborate each other or point in different directions; surveys may point to users’ awareness that various companies are generating large-scale behavioural surveillance data about them; and ethnographies can also be used to generate further information on participants’ uses of various devices to corroborate survey results.

Third, however, there can also be contestation: the various methods clearly also produce diverging perspectives, and the uncovering of such contradictions may point to the limitations of the methodologies themselves. For instance, the participant responses to surveys and interviews may not accurately reflect participants’ actual behaviours.

In this sense, working across methods and datasets also means working across the problems of fragmentation; scaling up the explanatory scope means reconfiguring the research project’s logic of anticipation.