Since its opening in 2009, the Dutch national organization for suicide prevention (113) has seen an annual increase in the number of people struggling with suicidal thoughts and behavior, with almost 140,000 chat and telephone conversations held in 2021. These conversations aim to keep clients safe from suicide and to help them seek help for their problems. Motivational interviewing (MI) is a crucial anchor in the (chat-)conversations for guiding the counseling process because this intervention can bring about demonstrable behavioral change. MI is a counseling style that helps people change their behavior by resolving ambivalence about unresolved or contradictory feelings. Earlier research has shown that counselors at 113 applied MI techniques consistently during chat conversations, but they could not strategically deploy MI techniques to elicit enough change talk from clients to change their behavior intrinsically.
Artificial intelligence (AI) offers enormous potential to analyze datasets to gain insights into developing and improving evidence-based interventions such as MI. In his research, Mathijs Pellemans investigated how to support counselors at 113 in applying MI during counseling sessions using AI. For this case study, Mathijs used a dataset of 253 actual chat conversations (approximately 24,000 chat messages) to classify client and counselor behavior based on language use. A trained AI model predicting client and counselor behavior achieved an accuracy of 72% (Cohen's kappa = 0.69) for classifying counselor behavior on 17 behavioral codes. The model scored 90% accuracy (Cohen's kappa = 0.66) on the classification of whether or not a counselor's statement elicits change talk. Moreover, for the effective elicitation of clients' motivational statements, it was essential to emphasize the client's autonomy and ask positive, open questions. In the future, such direct feedback can support counselors on the chat line with motivational interviewing.