SociRec: A user-centric, hybrid system for recommending social services to users
DescriptionThe design of systems that facilitate access to government-provided social services such as education, welfare, and job assistance by persons in need of care has received significant attention over the years. Unfortunately, these tools are not explicitly designed from a user-centered perspective. For example, when searching for social services, users may be required to precisely know the exact query keywords to initiate the search process, which may be a challenging task as they may struggle articulating their complex information needs. Moreover, services provided in response to a user's inquiry may not be specifically oriented to them based on unique attributes such as age, background, and job status. Thus, it is imperative that system designers approach both interface design and algorithmic implementation from a user-centric standpoint. We anticipate that doing so would lessen the effort required from users as they seek to access relevant social services. With this in mind, we argue for the need to enhance the information seeking process through the design of a recommendation system that explicitly targets users requesting access to social services. Specifically, we propose the design of a hybrid, user-centered recommender system. The model's recommendation algorithm leverages a wide and deep neural network architecture which simultaneously accounts for descriptive information about a social service and attributes unique to a user.