Accepted Paper:

The algorithm/data thing: Four metaphors for unpacking new methods in official statistics  

Author:

Baki Cakici (IT University of Copenhagen)

Paper short abstract:

I use four definitions of the thing—as object, as assembly, as superhero, as assimilating parasite—to investigate the entanglements of algorithms and data. Based on ethnographic research at national statistical institutes, I argue that the "algorithm/data thing" reshapes what counts as evidence.

Paper long abstract:

Algorithms and data are deeply intertwined. Data are gathered and stored using algorithms, and algorithms operate on data to produce results. To move away from the technical definitions of these two concepts, and to investigate them as objects of sociological interest, I propose understanding them together using the term "algorithm/data thing". I argue that we can use four different definitions of the thing—as object, as assembly, as superhero, and as assimilating parasite—to investigate the entanglements of algorithms and data. With these definitions, I extend Geoffrey Bowker's reading of Binder et al.'s proposal to consider alternative meanings of "thing", as in the Icelandic parliamentary institution "Althing", with two additional figures from popular culture.

I apply the algorithm/data thing to preliminary findings from a collaborative ethnographic research project spanning several European National Statistical Institutes, and I demonstrate how the algorithm/data thing reshapes what counts as evidence in new official statistics methods that use big data analytics and other experimental data sources. As a site of technical practice, as a mode of governance, as a key figure in contemporary discourse, and as a mechanism of quantification, the algorithm/data thing enacts specific types of evidence at the expense of others, and it has implications for how knowledge production is organised within national statistical institutes.

Panel T045
New Collective Practices of Measurement, Monitoring and Evidence