This paper critically analyzes the reality that most research on recommender systems is focused on a small number of application areas such as e-commerce and media recommendations, and spends a lot of computational resources on developing complex models without user evaluation or practical application. It points out that the scientific, economic, and social value of the research is often unclear, and argues that case studies in which recommender systems contribute to social good (RS4Good) should be emphasized more. To this end, it presents cases in which recommender systems have been successfully applied to solving social problems, and emphasizes the need for a paradigm shift centered on humanistic collaboration and long-term user participation evaluation.
