Changing maladaptive cognitions in traumatized children
Using artificial intelligence analysing 10,000 sessions of trauma-focused cognitive behaviour therapy.
Project Manager
Project Members
- Jensen, Tine Kristin
- Kjuus, Sigrid
- Ormhaug, Silje Mørup
- Nora Braathu
Main objective
The primary objective is to identify treatment aspects and techniques that lead to reduction of maladaptive trauma-related cognitions and subsequent symptom improvement in TF-CBT. This contributes to increase our understanding of therapeutic resolution of PTSD. We will do so by identify and classify therapeutic techniques and children’s emotional engagement based on audiotaped sessions, and combine this with self-report data on alliance, maladaptive cognitions and symptoms of PTSD.
Method
We will conduct a naturalistic mixed methods observational study where qualitative observational data will be combined with quantitative data from independent psychologist specialists as well as quantitative self-report data from the children. The data sources are: 1) 10,000 audio files/transcriptions of TF-CBT sessions with 550 children and their therapists. 2) Self-reported data on maladaptive thoughts and PTSD symptoms from validated questionnaires. 3) Assessments from psychologist specialists regarding therapists’ adherence to TF-CBT components.
The audio recordings are transcribed into text, which is then analyzed using locally installed, open large language models. We perform automated coding according to a fixed codebook, which is verified through human coding.
All data processing takes place within Services for Sensitive Data (TSD), a secure research environment at the University of Oslo. The models are used solely as analytical tools and are not trained on the data.