Health Recommender Systems for Mental Health Promotion
Torkamaan, H. (2022). [PhD thesis, University of Duisburg-Essen].
PhD thesis by Helma Torkamaan
Abstract
Recommender systems are today an essential part of software applications used in everyday life and facilitate the decision-making process for users by personalizing the options from which they can choose. An emerging and rapidly growing application domain for these systems is health care, and the majority of research contributions related to health recommender systems are about preventive health care. However, some crucial areas in this domain have been mostly overlooked. One such area is mental health promotion, which, despite its critical importance, has a relatively negligible share in existing solutions and research. User stress and mood are fundamental concepts in preventive health care, and proper skills for coping with stress and improving mood are crucial for individual mental well-being, which is the central theme of this dissertation. Health recommender systems have high potential benefits in personalizing health-related recommendations and, especially, engaging users in behavior change processes. A health recommender system for health promotion and behavior change is a holistic system that, ideally, uses techniques from ubiquitous computing to provide pervasive health. Building a health recommender system, therefore, is a multidisciplinary effort that engages various areas, which we summarize as tracking, interacting, and personalizing components, and address them in this dissertation regarding our recommendation domain, stress reduction. In particular, we discuss three major contributions to the problem of building health recommender systems for stress reduction and mood improvement: (1) establishing proper ways to track user mood; (2) building one of the first interactive mobile health recommender system research platforms and providing an extensive holistic dataset for flexible investigation of health recommender systems in the future; and (3) developing dynamic mood and health-aware, user-engaging algorithms and carefully comparing the performance and characteristics of the presented techniques. These contributions were the result of various mixed and longitudinal user studies which engaged with more than 2,500 users. This dissertation brings together for the first time various aspects of user decision-making - such as explicit short-term preferences, health needs, and long-term goals - for a holistic health-aware recommender system. By thoroughly discussing various components, this dissertation presents a roadmap for building health recommender systems, and interactive, mood-aware, and mental health-promoting systems in the future.
