After hating the book of the moment, Shoshan Zuboff’s much-praised Surveillance Capitalism, perhaps it underlines my contrariness if I tell you how much I loved my latest read, a book about classification. It was Sorting Things Out by Geoffrey Bowker and Susan Star, quite old now (1999). I can’t remember how I stumbled across it, but it absolutely speaks to my preoccupation with the fact that we see what we count & not the other way around.
The book investigates the confluence of social organisation, ethics and technologies of record-keeping as manifest in the establishment of systems of classification and standards. The examples it uses are medical systems such as diagnostic manuals, but the arguments apply more broadly. The point it makes about the role of record keeping technologies reminded me of a terrific book I read last year, Accounting for Slavery by Caitlin Rosenthal, which explored the role of commercially produced record books in the managerialism of large slave plantations in the US. The argument that a classification system lends the authority of something seemingly tehnocratic to highly political or ethical choices echoes Tom Stapleford’s wonderful book The Cost of Living in America.
As Bowker and Star point out, classification systems shape people’s behaviour. They come to seem like natural rather than constructed objects. They also fix perceptions of social relations, as a classification framework or set of standards, “[M]akes a certain set of discoveries, which validate its own framewor, much more likely than an alternative set outside the framework.” To switch frameworks requires overcoming a bootstrapping problem – you can’t demonstrate that a new one is superior because you don’t yet have the units of data on which it relies. People can’t see what they take for granted until there is an alternative version not taking the same things for granted.
And, although this book was written early in the internet era, the authors note that “Software is frozen organisational and policy discourse” – as we are learning with the burgeoning debate about algorithmic accountability. The essential ambiguity of politics is impossible to embed in code. The big data and AI era will force some of the fudged issues into the open.