| dc.description.abstract | Together with Sørlandet sykehus helseforetak and their team of specialists from the Selvmordsforebyggende nettverk, this thesis investigates the use of artificial intelligence (AI) to play the role of a patient in a virtual reality (VR) simulation. This represents a new application of the VR headsets already employed by the hospital for simulation training using 360-degree video. The thesis examines whether AI has reached a point where it can mimic human behavior convincingly enough to be an effective training tool.
Simulation training enables healthcare personnel to confront challenging situations in a safe environment. Learning from these experiences prepares them for real-life situations. Currently, the hospital employs various simulation methods, with staff role-play and 360-degree videos being the most common.
To create a realistic simulation for this thesis, data were collected from the specialists at the Selvmordsforebyggende nettverk using a mixed-methods approach. Initial semi-structured group interviews provided insights into the topic of their work, also gathering data from the first prototype. Further data on user satisfaction with the final version were collected through individual interviews and a survey. The simulation was developed using the following main components: Inworld, Unreal Engine, and MetaHumans.
The final version of the simulation offers cost and time efficiency, as well as replayability due to the AI's ability to provide varied responses. However, the AI in the current version cannot perfectly mimic a patient due to some unnatural facial expressions and occasionally unrealistic, lengthy answers.
Future research will include continued testing and adjustment of the AI for better results. Additionally, there will be exploration into other fields where this AI-supported simulation could be applied. | |