Working for the algorithms
As I leave the library, I reflect on the experience of working as part of a computerised system.
In my last day working for the library service,* the pressure was on to finish one of the regular administrative processes designed to ensure that the service’s stock of books is put to best use. This consisted of fetching a list of ‘dead’ stock, which is to say books that hadn’t been borrowed for a while. These copies needed to be looked at, and a decision made on whether they should be kept. Within each stock category, as well as a list of the books to be looked at, there was also a minimum and maximum number to be removed. In general, the minimum number was zero. Our stock management system has a similar process for identifying potentially ‘grubby’ stock for staff to investigate and consider removing. The ultimate decision always rests with the staff, and most removed stock is cycled back through the central stock team to libraries where it might be better used.
Dull though the above description is, this process got me thinking about the place of people in the modern working system. The algorithms driving our library management system have a pretty clear idea of what ought to happen in each library, and in theory the staff have a fairly mechanistic role of assessing the physical quality of the book and whether there is any reason to keep it in branch. Other than this, the only need for humans is that, as things stand, they are the best technology for taking things off shelves. This is especially the case in public areas like libraries, where other humans (i.e. borrowers) might disrupt automated shelf pickers. The actual role played by the staff, however, is not what one might imagine from a formal description of this mechanism. Rather than implementation of an algorithm, I would argue that our main role in this process was correction of the system.
To a library management system, all books are simply products to be judged by their figures. A history of the local village can be compared to the latest thriller or high profile celebrity bio, by the number of times they have each been checked out. Within each category the tolerances for number of borrowings can be adjusted, it’s true, so that it is not expected that the village history actually achieves the same number of issues as the thriller. But within a category the system will identify those books that are least loaned, and assume that these are the least important to the service. The role of staff consists largely in ignoring this simplistic judgement. I will start with the most dramatic example of this and then speak to its broader relevance.
The library I worked in has a large local history section, and every couple of weeks the local history society provide volunteers so that members of the community can drop by and ask questions about the area. A classic use for a local library, and one that still provides knowledge quite hard to come by in other ways: it is true that the Web has revolutionised historical research, but the more specific and local the knowledge you want the more likely it is to still be held in particular documents or simply in the minds of particular individuals. This kind of local history centre therefore remains a key access point for historical knowledge, and the published materials held by the library give a level of detail it would take a significant investment of time to reconstruct from scattered digital sources, assuming one could at all. In following the dead stock plan for this section, the role of staff is to simply ignore the management system’s recommendations. The value these books offer is almost impossible to measure through borrowings, because they are largely used for reference. Even if one could track their reference use in the system, one would not be justified in using this as a guide to remove stock, as the knowledge stored in that small area of a local library is highly unlikely to be replicated in any other such convenient location. It’s value lies as much in its availability as in its level of use. The library management system recommends that at least c.500 books should be removed from this collection, and a maximum of c.700. The collection itself numbers c.900. In other words, a few iterations of the algorithm without human intervention would see this resource rendered pretty near useless.
You may think that this example is isolated, a case where one could simply rewrite the system to ignore this section and be done with it. But actually it points up the paradox at the heart of this approach to stock management: it assumes that what is popular now is what a library should keep, and that anything else can safely be moved along. This view can be applied quite straightforwardly to, say, books released as TV tie-ins or the biographies of only briefly famous individuals. Libraries do not have infinite space, and something must give way for new titles, so it makes sense that those items that date the most quickly needn’t be kept out of some desperate need to archive everything. Nonetheless, there are classics within every genre, books that contain great beauty and wisdom and should be available as a resource just as the local history collection is. Yesterday we ‘saved’ biographies, histories, travel guides and others that keep our offering valuable, that allow the library to exist as a real service for people, not simply an alternative to Waterstones. You can wander in and peruse these books still, and take pleasure in them whether or not you borrow them.
And I suppose that brings me to my fundamental point: a library is not a shop. A bookshop is essentially aimed at sales, and so can use sales as a metric for all its stock. But the use of issues as a metric across stock fails to understand that a library is essentially a resource, and that the reading of books on site without borrowing them is not a negative in the same way it is for a bookshop. There are children who regularly use the library as a base for their homework, conducting research online and supporting it with the books on site, but perhaps mostly using the library because it is quiet and away from home. But if borrowings become the main way libraries judge themselves or their stock, we are saying that these children are not using the space at all, as their use is something that is not measured.
When judging public services, be very wary of using the metrics of the private sector. And next time your in a library, ask yourself why their most popular titles are grouped as ‘bestsellers’.
*I’m not going to name my former employer in this post. It is not written against them, but against a way of working that is increasingly common across sectors
