Understanding User Behavior in Carousel Recommendation Systems for Click Modeling and Learning to Rank

Carousel-based Recommender Systems, Learning to Rank, Click Models

Although carousels (also-known as multilists) have become the standard user interface for recommender systems in many domains (e-commerce, streaming services, etc.) replacing the ranked list, there are many unanswered questions and undeveloped areas when compared to the literature for ranked lists. This is due to two significant barriers: lack of public datasets and lack of eye tracking user studies of browsing behavior. Clicks, the standard feedback collected by recommender systems, are insufficient to understand the whole interaction process of a user with a recommender requiring system designers to make assumptions, especially on browsing behavior. Eye tracking provides a means to elucidate the process and test these assumptions.

In this extended abstract, the PhD project is outlined, which aims to address the open research questions in carousel recommender systems by: 1) improving our understanding of users' browsing behavior with carousels, 2) formulating a new click model based on the empirical evidence of users' behavior, and 3) proposing a learning to rank algorithm adapted to the carousel setting. For this purpose, we will carry out the first eye tracking user study within a carousel movie recommendation setting and make the resulting unique dataset of users' gaze and clicks publicly available.