“Rate My Therapist”: Automated Detection of Empathy in Drug and Alcohol Counseling via Speech and Language Processing

The primary data are 200 Motivational Interviewing (MI) sessions from a study on MI training methods with observer ratings of counselor empathy. Automatic Speech Recognition (ASR) was used to transcribe sessions, and the resulting words were used in a text-based predictive model of empathy..


At present, clinical practice and research in substance abuse treatment and mental health generally involves little to no direct evaluation of provider behavior. Psychotherapy research has long relied on the labor intensive and error-prone process of collecting observer ratings.

However, real-world training and service delivery demands are orders of magnitude larger than even our largest research studies. In many locales, the service delivery system and counselors are in place, but the training and quality assurance methodology (i.e., behavioral coding) are hopelessly mismatched.

The utilization of computational tools from speech and language processing present a technological solution that may ultimately scale up training, supervision, and quality assurance to match the service delivery need.

Bo Xiao,
Zac E. Imel ,
Panayiotis G. Georgiou,
David C. Atkins,
Shrikanth S. Narayanan