Virtual Standardized Patients for Cognitive Behavioral Therapy Training: Description of Platform Architecture

There is a significant need for novel technologies that allow cliniciansin-training to practice with an interactive virtual standardized patient (VSP; based on real-life patients). Building on previous successes, virtual reality, artificial intelligence (AI), and natural language processing (NLP) technologies are used to develop and test a robust web-based Virtual Insomnia PatientsTM (VIPs) platform. The AI-VIP responds strategically to input by utilizing a combination of expert VSP systems and deep learning techniques. The expert system uses the content collected from the Structured Clinical Interview for Sleep Disorders. Our VIPs involve a hybrid design process that mixes Agile and User-Centered iterative approaches with 3 main components: 1) realistic and artificially intelligent avatars for interacting with training clinicians; 2) a front-end system that implements multiple virtual avatars of varying race, ethnicity, and genders built using the Unity game engine; 3) back-end system that handles data storage, automates diagnostic accuracy and therapist fidelity measures to provide real-time comparison and feedback. The realtime feedback system employs natural language processing of a trainee's textual interactions with the VIP using computational models from the language used by real-life trained therapists. The VIP platform involves a universal storage language for the VIP dialog and symptoms that is updatable by trained clinicians, as well as a standardized 3D model system for the avatars allowing the selection of animations to match symptoms. VIPs will increase the availability of treatment, improving service members' psychosocial functioning, psychological and physical health, and overall fitness and decreasing accidents and military expenses.