Projekte

Alcohol use disorder (AUD) places a huge burden on those affected, their families, and the general community. Repeated cue exposure training is an effective adjunctive treatment method. However, it is rarely carried out due to limited therapists’ resources, and when it is, the sessions lack ecological validity as they seldom take place in the individual’s typical drinking environment. The development of digital assistance technologies may relieve therapy providers from workload and allow patients better and faster access to treatment in an ecological environment. One possibility is the development of virtual reality therapy assistants (VRTAs) based on chatbots that are enabled by current generative artificial intelligence (AI) techniques, especially large language models (LLM). Such assistants may guide patients
through adjunctive psychotherapeutic exercises and homework assignments. In this technical and clinical feasibility project, we aim to develop prototypes of VRTAs that train patients with AUD in successful exposures to drinking cues and related contexts by narratively and visually leading them through situations in which they regularly tend to use alcohol. The VRTAs will be based on current LLM technologies (i.e. Llama 3.1, GPT-4 from OpenAI,...) that have been modified and adapted to behave in strict accordance with best therapeutic practice and follow current state-of-the-art treatment manuals for cue exposure training in AUD. Experts by experience (patients) and clinical practitioners specialized in addiction medicine will test and evaluate the VRTAs. They will assess the feasibility of their use based on technical performance, user experience, and influence on clinically relevant outcomes.