Consider a catalog for a small public library that features a way to sort search results by popularity. There are several ways to measure “popularity” of a book: circulations, hold requests, click-throughs in the catalog, downloads, patron-supplied ratings, place on bestseller lists, and so forth.
But let’s do a little thought experiment: let’s use a random number generator to calculate popularity.
However, the results will need to be plausible. It won’t do to have the catalog assert that the latest J.D. Robb book is gathering dust in the stacks. Conversely, the copy of 1959 edition of The geology and paleontology of the Elk Mountain and Tabernacle Butte area, Wyoming that was given to the library right after the last weeding is never going to be a doorbuster.
So let’s be clever and ensure that the 500 most circulated titles in the collection retain their expected popularity rating. Let’s also leave books that have never circulated alone in their dark corners, as well as those that have no cover images available. The rest, we leave to the tender mercies of the RNG.
What will happen? If patrons use the catalog’s popularity rankings, if they trust them — or at least are more likely to look at whatever shows up near the top of search results — we might expect that the titles with an artificial bump from the random number generator will circulate just a bit more often.
Of course, testing that hypothesis by letting a RNG skew search results in a real library catalog would be unethical.
But if one were clever enough to be subtle in one’s use of the RNG, the patrons would have a hard time figuring out that something was amiss. From the user’s point of view, a sufficiently advanced search engine is indistinguishable from a black box.
This suggests some interesting possibilities for the Evil Librarian of Evil:
- Some manual tweaks: after all, everybody really ought to read $BESTBOOK. (We won’t mention that it was written by the ELE’s nephew.)
- Automatic personalization of search results. Does geolocation show that the patron’s IP address is on the wrong side of the tracks? Titles with a lower reading level just got more popular!
- Has the patron logged in to the catalog? Personalization just got better! Let’s check the patron’s gender and tune accordingly!
Don’t be the ELE.
But as you work to improve library catalogs… take care not to become the ELE by accident.
How to build an evil library catalog by Galen Charlton is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
I’m sure that some ethical (or at least ‘not-unethical’) mischief can be achieved.
I’m not convinced even of how ethical the popularity rankings options are in search results. Sure, it’s often up to a user to choose to sort in that way, but what about those titles displayed on the home page because they’re popular? In many cases those books will already be unethically at the top of the lists through dodgy practice of popular book lists and aggressive marketing. For a library to offer a ‘most popular’ is aggravating this problem.
I’d be all up for introducing some randomly generated results – to draw particular attention to books that aren’t checked out so often and experiment to see how that affects issues/whatever.
Indeed, ranking by popularity can be ethically tricky even without misrepresenting what it’s doing by randomly generating the rating. As you point out, basing it on bestseller lists and marketing data would allow the results to be gamed, either directly or indirectly. Even if the library only uses endogenous metrics such as circulation frequency, privileging the “usual suspects” can be problematic.
Much depends on the user’s purpose for sorting by popularity. Some folks may simply want to keep up with the zeitgeist and read what others are reading — and so be it. However, users who are using popularity as an indicator of quality can easily be misled. And of course, if popularity ranking actually does influence what patrons choose to read [citation needed?], positive feedback loops could narrow a community’s reading, bury books by minority authors, and make collection development policies less equitable.
I agree that there is a place for randomization when highlighting books in the catalog, but such displays should clearly indicate what’s going on.