A reinforcement learning approach to improve user achievement of health-related goals

Abstract

The demand and interest for personalized, efficient, and inexpensive healthcare solutions has significantly increased over the last decade to overcome the major limitations of existing traditional healthcare approaches. This new trend relies on the definition of intelligentmechanisms that can persuade the end-user to achieve health-related outcomes and ultimately improve his health condition and well-being. In this sense, the work here proposed explores a Multi-Agent System composed by personal agents that follow user preferences and a coaching agent which relies on a reinforcement learning approach to identify the most impactful messages to persuade a certain agent to follow established health-related goals. To validate the proposed system, a set of simulations were performed considering different types of persuasive messages and we were able to identify the most adequate sequence ofmessages that can persuade different users to achieve health-related goals based on their preferences.