Literature Case Study
Goodreads is a free book-focused social networking and cataloging platform that is owned by Amazon. On the site, users can create lists of books that they have or want to read, write reviews and rate books. In return, the site provides book recommendations, reading challenges and shares community ratings and reviews.
In 2011, Goodreads adopted a machine learning algorithmic recommendation technology to provide personalized recommendations, through book ratings and trends (Chandler, 2011). The ML technology also improved Goodreads targeted advertisements (Chandler, 2011). Per Goodreads, their “recommendation engine” uses algorithms to analyze 20 billion data points, mapping connections between books and readers (Chung, 2011). The program proposes recommendations based on what users have put on their digital bookshelves, and will suggest books based on similar genres and themes that other users with comparable bookshelves have enjoyed. To improve recommendations, users can rate more books, indicate books they are not interested in, tag books and select their favourite genres.
Despite user inputs to help Goodreads’ ML improve, a criticism of Goodreads’ system is that many users feel that the recommendations are not sufficiently personalized. Rather, it seems that the algorithms suggests books based off of the genres that are already on users’ bookshelves instead of helping users explore new genres, themes or moods (Stochl, 2021). Based on this criticism, it appears that the ML-generated recommendations would affect users’ behaviour by creating a feedback loop. In this case, a reader might highly rates a certain genre of book, so the system recommends more books within this genre, and when a user positively reacts to a recommendation, the process will repeat. The issue is that readers have diverse interests, and may not like all books within a genre, or they may wish to read a variety of genres. The recommendations simply provide users with suggestions that are thematically close to previous reads, and do not encourage readers to expand their literary interests, which may lead to the next great book. Further, books (and authors) that do not fit into distinct or popular genres may not receive attention, thus penalizing titles that push or fall outside of the boundaries of mainstream literature.
Many users enjoy Goodreads for its cataloguing feature, however the recommender system provides value added to users, as well as Amazon, the parent company. By knowing which books a user has on their bookshelves, Amazon is able to tailor advertisements in the same way that Goodreads can provide recommendations for the next read. For example, Goodreads creates reading lists based on users’ interactions with features. This includes a monthly list of the most anticipated new books based on the number of users that added a book to the their “Want to Read” shelf (Cybil, 2022). Advertisements on the Goodreads platform include books published by Amazon Publishing. These advertisements may generate more interaction with the promoted books, which the platform’s algorithm will in turn measure as interest. If users or the system add promoted books to a reading list, it will be promoted for a second time. The more a book shows up on reading lists and as a recommendation, the more users will see it, ultimately impacting book sales.
