Substance use disorders can be effectively treated by evidence-based counseling approaches. Indeed, millions of Americans receive counseling for each year. However, there is no scalable method for sustaining treatment quality. In community settings supervision rarely occurs, and does not include performance-based feedback. Methods for evaluating treatment fidelity were designed in research settings and rely on human-intensive processes that are expensive and time-consuming (e.g., human raters evaluate recordings).
Counseling is essentially a conversation – the data is spoken language. Advances in natural language processing have made it possible to explore the content of psychotherapy at a scale and level of detail that has not been possible before. I will describe my group’s current work in this area, in particular focusing on how new tools can be used to evaluate the content of
As a result of attending this session, participants will be able to:
1) Introduce attendees to basic concepts related to machine learning in mental health. Describe current research on the evaluation of psychotherapy data with machine learning tools; and,
2) Describe new findings related to predicting treatment outcomes directly from session recordings.
