Modelling therapeutic alliance using a user-aware explainable embodied conversational agent to promote treatment adherence

Non-adherence to a treatment plan recommended by the therapist is a key cause of the increasing rate of chronic medical conditions globally. The therapist-patient therapeutic alliance is regarded as a successful intervention and a good predictor of treatment adherence. Similar to the human scenario, embodied conversational agents (ECAs) showed evidence of their ability to build an agent-patient therapeutic alliance, which motivates the effort to advance ECAs as a potential solution to improve treatment adherence and consequently the health outcome. Building therapeutic alliance implies the need for a positive environment where the ECA and the patient can share their knowledge and discuss their goals, preferences and tasks towards building a shared plan, which is commonly done using explanations. However, explainable agents commonly rely on their own knowledge and goals in providing explanations, rather than the beliefs, plans or goals of the user. It is not clear whether such explanations, in individual-specific contexts such as personal health assistance, are perceived by the user as relevant in decision-making towards their own behavior change. Therefore, in this research, we are developing a user-aware explainable ECA by embedding the cognitive agent architecture with a user model, explanation engine and modified planner to implement the concept of SharedPlans. The developed agent will be deployed and evaluated with real patients and the therapeutic alliance will be measured using standard measurements.