Users have an actual conversation with Barry, and he responds in real time.
The Western Australian Future Health Research and Innovation Fund (FHRI) has awarded a $499,660 grant for the further research and development of an immersive, artificial intelligence (AI)-driven, virtual-reality-enhanced computer simulation program to train frontline healthcare workers in de-escalating aggression in patients.
"Barry" is an AI computer-based model that allows clinicians and students to verbally interact with him in a hospital setting. Users have an actual conversation with Barry, and he responds in real time.
Given the use of AI, Barry's response to each verbal interaction is entirely unique, even if users provide the same wording more than once.
The training simulation incorporates emotive facial animations that align with real-time verbal responses, and Barry's anger levels will either escalate or de-escalate based on the empathy displayed from the user.
The unique responses given by Barry are based on upfront prompting from the development team, outlining the gender, race and age of the user undertaking the training, which can be customised to suit different user profiles.
"We were able to develop a basic proof-of-concept from smaller Phase 1 funding, what this Phase 2 funding will mean is that we can take this proof-of-concept through a full R&D cycle for scalable deployment within health systems," Project Lead Dr Brennen Mills said.
"Health services are crying out for a training solution to combat the issue of workplace aggression and violence. VR provides a flexible training solution, and when combined with AI will provide highly authentic practice exposure which is so important for development of de-escalation skills."
The funding will be used to enhance the AI-model better ensuring Barry's responses map aptly with input from the user. It will also enhance the emotive facial animation and speech pipeline to reduce lag time between responses, as well as create a secondary AI that will track Barry's anger level and tailor responses to match, including opportunities for Barry to turn physically violent.
Additional work will also address deployment considerations ensuring the application can be implemented within health services, as well as examination of Barry's responses adapting based on user's demographic characteristics.
