This week in RCDH, we focused our reading and discussion around algorithms. As I mentioned last week, the topic felt kind of transitional for me—databases, archives, and metadata blend together fairly well (for me, at least), and they’re not topics that feel overwhelmingly technological for people. Whether or not we work on the back ends of those kinds of projects, the concepts themselves are not immediately intimidating, I don’t think.

That changes a bit with the shift to algorithms, which have a more machinic flavor. Whether that’s actually the case is something I was thinking about in class, and it’s persisting with me this morning. One of the things that I said last night is that, at heart, algorithms are simply procedures, and we spent a healthy chunk of our time trying to put that into practice.

Last week, we used a couple of pages from an old MLA job list, and brainstormed as complete a catalog of metadata as we could. Before class yesterday, I took that list and converted it into “variables,” such that Rank, for example, became a class with potential values like Assistant, Associate, Full, Open, Fixed Term, etc. I brought copies of that list of variables to class with me, and had the students write “programs” for evaluating job advertisements. We didn’t focus too heavily on programming languages or anything—I wanted to keep it as light as possible. The goal of the program was to generate a score S(n) for each of the sample ads I brought with me, and for the most part, the programs were simply a series of if-then-else steps that bumped up or down S based on a range of factors (geography, specialization, etc.). The students then swapped programs, and used them to score several ads.

Some of the obvious objections emerged in discussion—I didn’t provide any instructions about the value of S other than it should start at zero, and so we ended up with scores that ranged from single to triple digits. I didn’t give them copies of the ads ahead of time, so some of the variables they used weren’t especially relevant. I asked the students to execute the programs “as is,” which restricted what is normally a more recursive process. I probably should have been more careful about choosing information-rich ads and pruning down the variables to minimize wasted effort, but otherwise, it was an interesting exercise. It also turned us to some conversation about how the whole ecology of the process would need to change in order to implement this program for all to use, how search committees engage in their own procedures as they read materials, etc.

Maybe the most interesting thing for me about the exercise was the process of translating what is often a personal, value- and preference-oriented practice into algorithms. One of the pieces we read was Tarleton Gillespie’s The Relevance of Algorithms, an essay available online and also recently published in Media Technologies: Essays on Communication, Materiality, and Society, fresh off the press from MIT. Gillespie’s essay makes several points that really stuck out for me this week; among them is the idea that algorithms instantiate certain kinds of “knowledge logics,” such that it makes as much sense to study them sociologically as it does to think of them as technologies. He locates a “fundamental paradox” at the heart of algorithms:

Algorithmic objectivity is an important claim for a provider, particularly for algorithms that serve up vital and volatile information for public consumption. Articulating the algorithm as a distinctly technical intervention helps an information provider answer charges of bias, error, and manipulation. At the same time, as can be seen with Google’s PageRank, there is a sociopolitical value in highlighting the populism of the criteria the algorithm uses. To claim that an algorithm is a democratic proxy for the web-wide collective opinion of a particular website lends it authority. And there is commercial value in claiming that the algorithm returns “better” results than its competitors, which posits customer satisfaction over some notion of accuracy (van Couvering 2007). In examining the articulation of an algorithm, we should pay particular attention to how this tension between technically-assured neutrality and the social flavor of the assessment being made is managed — and, sometimes, where it breaks down (182).

It’s not simply that algorithms represent the operationalization or reification of what are social processes—there are also sociopolitical consequences to algorithms that make them our interlocutors and audiences, that place them among our available means of persuasion, and that make them rhetorical ecologies of their own. These are points that are obvious to some of you who might read this, but I don’t think rhetorical studies has quite yet gotten on board with the relationship between algorithms and rhetoric. I’m only just now articulating it for myself, to be fair.

One thing that occurred to me last night was an interesting parallel between Gillespie’s characterization and the definition of myths that Roland Barthes offers up in Mythologies. For Barthes, myths are a particular form of signification, an operation that takes the historical or the contingent and transforms it into Nature:

When it becomes form, the meaning leaves its contingency behind; it empties itself, it becomes impoverished, history evaporates, only the letter remains. There is here a paradoxical permutation in the reading operations, an abnormal regression from meaning to form, from the linguistic sign to the mythical signifier….The meaning will be for the form like an instantaneous reserve of history, a timed richness, which it is possible to call and dismiss in a sort of rapid alteration… (227)

I don’t know that I can convey to you how fascinating it’s been to read the first several pages of “Myth Today” in Barthes’ book, substituting “algorithm” for myth—in some places, the stretch is more casuistic than others, but by and large, it works. And this parallel might extend productively to Barthes’ thoughts about how to read myths: there are those of us who simply accept the results of algorithms as they are, those of us who understand their distortions, and then perhaps those who “focus on the {algorithm} as on an inextricable whole made of meaning {social flavor} and form {technological procedure},” producing an “ambiguous signification” that foregrounds Gillespie’s paradox.

There are strong parallels as well between a Barthesian “reader of algorithms” and the kind of computational literacy that Annette Vee advocates or the procedural literacy suggested by Lisa Gye:

procedural literacy entails learning, and thus being able to recognise, the procedures that enable algorithms and hence software to weave their magic but it is also a more fundamental literacy which takes into account of a range of human interactions. It allows us to model knowledge and to see the world as a system of interconnected parts.

It makes a great deal of sense to me to think of algorithms as myths in the Barthesian sense—and I wonder to what degree this parallel offers a more theoretically-inclined and/or humanities-friendly way of thinking about the issue of whether or not programming/coding belongs in our curricular core. In several ways, the arguments for the centrality of rhetoric are the same ones we might make for the importance of algorithms. I’m not sure that I’d completely recognized (or articulated) that resonance before now.

One last item. I want to write too about the other angle that our readings took last night, but I think I’ll save those for another post. I should note, though, that alongside several essays, we also read Stephen Ramsay’s Reading Machines, a book that makes much of McGann and Samuels’ notion of “deformance” or deformation:

In one sense, deformation is the only rational response to complexity. Nearly all deformative procedures (which include outline, paraphrase, translation, and even genre description) are intended to alleviate some difficulty…All textual entities allow for deformation, and given that interpretation occurs amid a textual field that is by nature complex, polysemic, and multi-referential, one might say that most entities require it. Seen in this light, deformation is simply a part of our permanent capacity for sense-making (48).

Are you ready to have your mind blown? According to Barthes, “The relation which unites the concept of the myth to its meaning is essentially a relation of deformation” (232).

Boom.

(And yes, I think I have the topic/angle for the next essay I’m going to write.)