Orientation Services and Environments Librarian, Associate Professor
University of Illinois at Urbana-Champaign
With research funding from the University of Illinois Campus Research Board, a personalized account-based recommender was developed in the University Library’s mobile app interface. The recommender system (RS) was derived from data mining topic clusters of items that are checked out together. Using the library mobile RS as a prompt to understand student preferences for personalized account-based RS, structured interviews were undertaken and analyzed thematically to determine RS features and functionality desired. In the interviews, students described their perceptions of RS, together with features and functionality desired. Students indicated that they desired data stewardship and sharing levels, which provided valuable input into matters of system transparency pertaining to recommendations derived algorithmically. An unexpected finding from students was growing unease with aspects of surveillance capitalism . One student referred to YouTube as an example of a service that did not work the way she wanted it to and noted that “…frequently YouTube doesn’t work so good because it gives you a recommendation based on one thing you did. It should be based more on a frequently searched thing. Recommendations are sending you things you already are interested in, which might not show you newer things and that is not really a good way to learn.” Students also indicated that they did not like the fact that commerce seems to drive recommenders, for example, “…on the Internet, you might be interested in finding information about something but not want to buy.” Another student took a measured approach to this problem, noting “…when they [recommenders] are trying to sell something it feels predatory, but otherwise, it is good.” This presentation will explore in greater detail the need to safeguard student privacy despite the finding that students believed that there is a place for RS in academic library settings. Academic library recommenders can distinguish themselves from commercial recommenders in several ways, including increased transparency beyond what is available in commercial systems, and by attending to the level of student privacy desired as a system design issue.