Transforming motivational interviewing into a computable model for automated patient diabetic counseling.

Project Abstract The incidence of type 2 diabetes mellitus (T2DM) is staggeringly high in the United States, and it is expected to exponentially increase in the next few decades. Regulating health behavior that leads to T2DM is a key factor that can slow the down prevalence of T2DM and decrease any complications resulting from T2DM.

Motivational Interviewing (MI) is one of the evidence-based approaches that providers can wield before considering more expensive and invasive methods. MI is a counseling technique, involving empathy and evoking “change talk” by the patient, that aims to resolve the ambivalence that prevents clients from realizing personal

goals. Together with the patient, the MI-trained provider can illicit the patients' hesitancy of unhealthy behaviors that increase the severity of T2DM and eventually improve the patients' health outcomes and lifestyle. Research has shown that MI is effective in modifying some of the health behaviors that relate to T2DM severity (e.g.,

regulation of sugar-sweetened beverages consumption and promoting physical activity). However, behavior change counseling is rarely implemented in clinical practice. The use of speech interfaces in devices have been increasing used by consumers over the last few years. Coupled with advancements in intelligence-based dialogue methods, software agents can automate dialogue

exchanges between machine and users, and could engage users using human language and conversational turns. If we could automate the counseling experience through a speech-based dialogue system we could delegate the task counseling task and overcome some of the challenges to implement MI – training, consistency,

reimbursements, and lack of time. The researchers presume that we can model and emulate the MI counseling method for machines to enact conversations to evoke patients' ambivalence of harmful health behaviors surrounding sugar-sweetened beverages and encourage light physical activity. In addition, we posit that motivational interviewing experts and

end users who have diabetes or pre-diabetes can positively assess the usability of our software agent (“app”), code named “TROI”, to preform MI counseling for our use-cases. This project aims to develop comprehensive ontology models (i.e., the knowledge base for machine interaction) for motivational interviewing for counseling diabetic-related behavior. This will involve

conducting simulation of MI counseling to help us analyze and understand the method to translate it to a computer-based model (ontology) and integrate that model towards the development of “TROI”. Lastly, we intend to evaluate the automated counseling experience by the ontology-based computer agent. This will involve

recruiting MI experts to assess the integrity of MI counseling and the recruitment of diabetic patients to assess their receptiveness of the tool.