This thesis aims to analyse the behaviour of an “analogue search engine”, created by the romance novel community. The existing commercial recommendation systems for romance novels are found lacking, as they fail to fill the requirements of the user group. The user group fills this need by creating separate communities, where they use their expertise in the genre to create a more accurate recommendation system.
Using ethnographic research over a long period of time allowed the behaviours and habits of the group to be catalogued. Using that framework and interaction design practises, several design proposals were created showing how the findings can be used to add nuance to recommendation system interfaces.
Here's the abstract: