Reading :: The Constitution of Algorithms

The Constitution of Algorithms
By Florian Jaton

In this book, Jaton uses a Latourean approach to examine how algorithms are constituted. I found the introduction to be really interesting -- the case studies less so.

Let me set the scene, since it may be affecting how I read the book. I actually read this book a few months ago, in PDF format, and added comments to it in Google Drive. Since then, UT has announced that we no longer have unlimited storage space on GDrive, and it turns out that I was far over the new quota. So I have moved all of my PDFs to my UT space on Box.com, which is unlimited—for now. Unfortunately, although the comments transferred over to Box, I can only see them by mousing over the page—there's no visual indicator that the page has a comment until I happen to mouse over it. That makes writing this review more difficult.

Fortunately, most of the comments are in Chapter 1, which—to my mind at least—is more valuable than the case studies that support it. In Chapter 1, Jaton considers the question of algorithms. From a Deleuzian standpoint, they have fluidity, swiftness, and distributivity, and can be characterized as devices: They circulate and link up sparse actants quickly (p.6). STS scholars have examined algorithms' workings and agency, but Jaton is more interested in how "unrelated entities (e.g., documents, people desires) ... come into contact to form, in the end, devices we may call 'algorithms'" (p.8). Jaton sees algorithms as durable, mobile, and carrying characteristics of other actants, just as Goody's graphical objects, Latour and Woolgar's inscriptions, and Dorothy Smith's documents (p.13). 

Jaton proposes to study these in terms of constitution rather than construction. The term constitution evokes a political settlement, and constitutions can be amended (p.17). To explore the question, Jaton introduces his ethnography of a computer science laboratory. 

Although I usually enjoy ethnographies, I confess that I was not that interested in this one. It reads a bit like Laboratory Life, but whereas LL provided momentously new insights to me, this ethnography mainly seemed to transfer those insights to a new activity. I did learn a lot about how algorithms are developed, but the insights felt small and incremental. As the author notes in the conclusion, "these empirical elements [from the case studies] might seem quite tenuous when compared with the ogre to  whom this book is explicitly addressed: algorithms and their growing contribution to the shaping of the collective world" (p.283). I agree. Although Chapter 1 is well worth reading, I don't think the case chapters measure up to it. 

Should you read this book? If you're interested in understanding where algorithms come from, or exploring them from an STS standpoint, definitely—but spend most of your time in Chapter 1.