The second ARITHMUS method workshop titled “Tracing Algorithmic Practices” took place on 2-3 November 2015 at Goldsmiths, University of London. It was organised to address the practical problem of doing a collaborative, multi-sited large-scale ethnographic study involving statisticians who work with algorithms of different scales and complexities. Fourteen researchers from various social science disciplines participated in the workshop. The workshop began with a keynote by Paul Dourish, titled “Algorithms Rather Than What? Algorithmic Culture in Context” on November 2.
During the workshop, we addressed algorithms as objects of practical attention, as they emerge in the work of different groups such as IT workers, computer scientists, and statisticians, among others. We argued that it is productive to see algorithms not as mathematical abstractions, but as practical resources used by these groups.
Three examples of outcomes from the workshop are:
- Agency in relation to algorithms and their mediation capacity is worth investigating. Both in terms of how agency is distributed and its implications for accountability, but also the accountability for how data are constituted prior to being processed by algorithms, rather than only for how algorithms function.
- The relationship between data representation and computational processing is often missing in the way algorithms and their consequences are discussed. There is a need to bring data and their constitutive effects back into the discussion, and to also investigate the historical and temporal entanglements of data and algorithms.
- The role played by the metaphor in the imaginings of the algorithm is striking. Metaphors have a bidirectional property. For example, not only are algorithms understood to operate similar to the behaviour of ants, but also that algorithm experts become experts on the behaviour of ants, and afterwards read algorithms into the behaviour of ants. Paying attention to what sort of work metaphors do and why is important, as not all metaphors work on the same principles or for the same purposes.