This short post started as a Twitter thread but I thought I’d drop it in here for future reference.


I recently got a real kick out of reading Adam Burke’s Occluded Algorithms that got me thinking in more concrete terms about how popular uses of the term algorithm are deployed relative to their use in computer programming. It builds on some previous work of Seaver (2017) which I’ve also briefly written about here.

It has become quite common to see news stories that talk about Facebook’s news feed algorithm, or Google’s ranking algorithm, as things that shape our experience of using computational tools. On the other hand the word algorithm has been used for decades to describe discrete computer processes defined in code that take a set of input and generate a set of output. A large part of the computer science research literature consists of discussions of these algorithms, and their application in various contexts..

The connection that Burke seems to make (at least for me) is that in both cases the term algorithm works from a particular perspective outside of computation, to talk about some complexity that is inside. Algorithm internals are often complex and difficult to understand. He uses the example of Python’s list sort which makes a sort algorithm available to the user. On the one hand this empowering because it offers this algorithm to the computer programmer for them to use. But at the same time it encapsulates and hides its implementation and doesn’t require the programmer to completely understand what it is doing. The positiionality here is similar to when a user of Facebook takes the ordering of their news feed at face value.

Algorithms get packaged up in modules and libraries where they can be orchestrated together. Much of what we call software development today involves the pragmatic swapping in & out and plugging together of these units of complexity.

Open Source offers the ability to open up and inspect the inner workings of algorithms. But this often involves a huge commitment to understand the complexity within, which is simply not feasible at the level of a system. But, thus far, for many eyes, it has proven feasible. Although it’s arguable that the eyes are getting strained.

But I guess this focused little piece by Burke really got my attention because of the way it knitted together theoretical ideas like agential cut (Barad, 2007), assemblage (Deleuze & Guattari, 1987), algorithms as culture (Seaver, 2017), technical objects (Simondon, 1958), and governmentality (Foucault, Davidson, & Burchell, 2008). This last point on governmentality and its connection to algorithms is something that Burke draws from Introna (2016), which is new to me.

I’m interested in how governmentality can provide a way of understanding a wide variety of theories and approaches to archival appraisal–especially in web archives. Apart from Mackenzie (2017) I haven’t seen a whole lot written about algorithms and governmentality. Since I’m mostly focused on web archives algorithms are an important topic. So if you find other connections please let me know.

References

Barad, K. (2007). Meeting the universe halfway: Quantum physics and the entanglement of matter and meaning. Duke University Press.

Deleuze, G., & Guattari, F. (1987). A thousand plateaus: Capitalism and schizophrenia. Bloomsbury Publishing.

Foucault, M., Davidson, A. I., & Burchell, G. (2008). The birth of biopolitics: Lectures at the Collège de France, 1978-1979. Springer.

Introna, L. D. (2016). Algorithms, governance, and governmentality: On governing academic writing. Science, Technology, & Human Values, 41(1), 17–49.

Mackenzie, A. (2017). Machine learners: Archaeology of a data practice. MIT Press.

Seaver, N. (2017). Algorithms as culture: Some tactics for the ethnography of algorithmic systems. Big Data & Society, 4(2).

Simondon, G. (1958). On the mode of existence of technical objects. Aubier, Editions Montaigne. Retrieved from https://english.duke.edu/uploads/assets/Simondon_MEOT_part_1.pdf