The limits to modelling dataflow networks: a conceptual synthesis

This paper offers an introduction to the task of studying dataflow networks as socio-technical systems. Dataflow networks are central to contemporary economies, and their modelling is important to many stakeholders. Participants seek to understand the risks and benefits of participation; those interested in governance seek to understand how systems operate and identify appropriate interventions. However, truly understanding these networks' structures-let alone how information propagates across them and how this gives rise to outcomes---cannot be performed from a technical point of view alone.Using the targeted advertising ecosystem as a case study, we discuss the issues around studying dataflow networks and identify three categories of limitations: information complexity, qualitative and indeterminate factors, and network opacity.To overcome these barriers and make the dynamics of datafied economies tractable, we argue for a combination of systems analysis with techniques from the social sciences. We contribute a synthesis of relevant tools and social theory, focusing on the processes by which data acquires value within socio-technical settings. We also put forward a general model which accounts for both market- and non-market-based datafied economies. Our approach is applicable to a wide number of fields where dataflow networks arise, including data exchanges in research settings, government data sharing initiatives, and surveillance ecosystems.