Anton Nijholt
The behavior of users in the digital world, such as online shopping or social media activity, is increasingly supported by personalized systems, such as recommender systems \(Ricci et al., 2015\) and personalized learning. Early work on personalized systems was mainly data-driven, based on behavioral data, such as ratings, likes, and purchases (e.g., Bell et al., 2007\). Although these systems are useful for both users and service providers, the main downside is the limited interpretability and explainability of the data. Such limitations in both interpretability and explainability translate in using data without understanding the root-cause of behaviors. Recent work has thus started to adopt a more theory-driven approach by including psychological theories and models to improve personalized systems (see for an overview; Graus and Ferwerda, 2019\). These systems take advantage of psychological theories/models, such as emotions (Tkalci ˇ c et al., 2013b; Tkal ˇ ci ˇ c and ˇ Ferwerda, 2018\), personality \(Ferwerda et al., 2017; Wu et al., 2018\), skills \(Ferwerda and Graus, 2018\), and culture \(Schedl et al., 2017\) to explain and predict behaviors of users. This allows for a deeper understanding of users' behavior, preferences, and needs, which in turn also lead to more generalizable results.