Identification of cultural conversations in therapy using natural language processing models.

Area of Science:

• Psychology
• Computer Science
• Clinical Psychology

Background:

• Traditional methods for assessing therapist multicultural competency rely on client self-reports and behavioral coding.
• These methods have limitations in providing therapists with systematic, scalable feedback on cultural discussions during sessions.

Purpose of the Study:

• To develop and evaluate natural language processing (NLP) models for automatically identifying cultural topics in psychotherapy.
• To explore the potential of NLP for providing therapists with rapid, data-driven feedback on multicultural competency.

Main Methods:

• Utilized a dataset of 103,170 labeled talk turns from 188 psychotherapy sessions.
• Developed and trained NLP models to recognize discussions related to cultural concerns within therapy conversations.

Main Results:

• The developed NLP models demonstrated high accuracy in recognizing cultural topics (F1-score = 70.0).
• Model performance showed strong correlation with human judgment (Spearman's ρ = 0.78, p < .001).

Conclusions:

• NLP models show significant promise for the automated identification of cultural discussions in psychotherapy.
• These findings suggest potential for innovative NLP-based tools to enhance therapist training and feedback in multicultural practice.