All data are local – notes from Loukissas (2019)

In his most recent book, designer Yanni Loukissas makes a simple yet compelling argument: all data are local. Loukissas’ book could be read as an extended rebuttal of the ‘myth of digital universalism’ (p.10), which prevails much of the rhetoric of digital data. Digital universalism, he argues, helps to uproot data so that they might be operationalised in new contexts ‘unburdened by ethical quandaries that might accompany their use, and free from concerns about their unintended consequences’ (p.125). By obscuring these local effects, new economies of data can be created and the market gains strength. Drawing on in-depth case studies from a range of data settings – including accessions data from the Arnold Arboretum, and the Zillow online real estate database – Loukissas details the importance of analysing data settings and data practices, rather than simply data sets. In this blog post I summarise the four supporting principles to Loukissas’ main argument that all data are local.

  • The first principle Loukissas explores is that all data have complex attachments to place. Data are not independent of their local substrate and carry with them traces of the conditions and values particular to their origin. While we might like to think of data as readily available for use, Loukissas details the huge amount of labour that goes into making data machine recognisable. In addition, data processes and practices within an institution change over time. Something as simple as the date and time structure used within an institution to label data can have profound effects on how datasets are constructed and used.
  • The second principle is that data are assembled from heterogenous sources. Loukissas focuses upon data infrastructures, or the ‘metacollection, which agglomerates digital resources from distributed sites of production’ (p.57). Drawing on the work of Bowker and Star (1999), Loukissas argues data infrastructures are not neutral and draws attention to how data are made to conform to a range of categories and structures. Rather than simply accepting data at face value, we need to ask critical questions that help to uncover the cultural, political and economic construction of data. These questions include: Who makes data and why? What does data mean in different cultures of collecting? What kinds of values and assumptions can data hold?
  • The third principle is that data and algorithms are entangled. While this is not a new insight, Loukissas explains how algorithms are trained on data sets and that this process fundamentally shapes their normative assumptions. Algorithms are ‘fragile, multiple, and situated in ad hoc practices of computational work’ (p.117) and therefore not easily separated from the data-sets that they are trained on and/or process. In this sense, any algorithm is not easily appropriated for other ends.
  • The fourth principle is that interfaces recontextualize data. Interfaces are key to universalising data as they enable data to be delocalised. Through the interface, data can be integrated from ‘anywhere’ and put to work ‘equally well everywhere’ (p.125). While interfaces provide a pragmatic solution to data use, they also help to establish subject positions that the users of data ‘are expected to adopt’ (p.126). As such interfaces grease the wheels of data generation and use. Loukissas argues data are never free, but are instead recontextualised through interfaces.

The lessons in All Data Are Local are useful to all who work with or on digital data – from social scientists to data scientists. As Loukissas demonstrates, data are only ever as reliable and useful as the processes and practices that are employed to manifest and use them. As such, data must be analysed in context and as a kind of ‘text’ or cultural expression ‘subject to interpretive examination’ (p.7). While data often enable people to make decisions ‘objectively’ and ‘from a distance’, Loukissas concludes with the contention that data should be seen as a ‘point of contact’ or an opportunity ‘to get closer, to learn to care about a subject or the people and places beyond data’ (p.196).


Bowker, G., & Star, S. (1999). Sorting Things Out: Classification and its Consequences. Cambridge, MA: The MIT Press.

Loukissas, Y. (2019). All Data Are Local: Thinking Critically in a Data-Driven Society. Cambridge, MA: The MIT Press.