Handling Conflicts in a Multi-Agent System for E-coaching

Abstract

According to the World Health Organization (WHO), in 2016, there were over 250,000 mobile health applications in major app stores [1]. However, most of the apps are only focused on one disorder, and there are very few apps (among the listed apps) that support multiple disorders simultaneously. Whereas in reality, often multiple disorders co-occur as one disorder leads to another disorder. For instance, diabetes is one of the substantial risk factors for heart diseases [2]; sleep disorders (insomnia and hypersomnia) co-occur with depression [3, 4]; substance abuse is associated with the depressive mood, anxiety disorder, and anti-social personality disorder [5]. We model a basic framework (the so-called MentalHeath-Care or MHC) of a e-coaching system where a coach coaches his/her clients via electronic media, i.e., internet to support numerous health problems simultane-ously and independently. The unique part of the system is that it would be a fully automated e-coaching system where there will be no human coach to coach the clients, but autonomous intelligent agents. We start the framework for two relatively close disorders, i.e., insomnia and depression which often co-occur. For two distinct disorders, we plan to have two autonomous agents (coaches) who will be situated in the same setting (smart-phone) and will share the same environment. Thus, MHS is built on the multi-agent system paradigm. The e-coaching system would be able to offer three types of coaching: 1) independent depression coaching, 2) independent insomnia coach-ing, and 3) simultaneous depression and insomnia coaching based on a client’s symptoms. However, to ensure the third kind of coaching the coaches require to work in a team and cooperate with each other as either of them might have incomplete or no knowledge of the other disorder. In order to have teamwork and cooperation, the coaches need to interact and communicate with each other. But, an important shortcoming of cooperation in a multi-agent system is that it may generate conflicts or disagreements between coaches on various aspects. The conflicts may regard to different knowledge, goals [6], plans, desires, and beliefs [7] of the coaches; shared resources and task interdependencies [8] between the coaches. In this thesis, we propose a conflict resolution mechanism for a multi-agent system based fully automated e-coaching system that offers coaching on different mental disorders, i.e., depression and insomnia.