Leveraging Large Language Models to Estimate Clinically Relevant Psychological Constructs in Psychotherapy...

Area of Science:

• Psychology
• Computational Linguistics
• Psychotherapy Research

Background:

• Quantitative measurement of psychopathology and treatment progress is challenging due to reliance on natural language.
• Previous linguistic measures of psychological distancing were context-insensitive, limiting their scope.
• Existing methods struggle to analyze subtle therapist language and its impact on client outcomes.

Purpose of the Study:

• To introduce and evaluate large language models (LLMs) for measuring linguistic distance in psychotherapy.
• To compare LLM-based measures with traditional word-counting techniques.
• To explore how therapist language influences client psychological distancing and outcomes.

Main Methods:

• Utilized appropriately prompted LLMs to assess linguistic distance in therapy transcripts.
• Compared LLM-derived linguistic distance measures against traditional word-counting methods.
• Analyzed both client and therapist language to understand their relationship with therapeutic progress and symptoms.

Main Results:

• LLMs provide a more nuanced and context-sensitive assessment of linguistic distance compared to traditional methods.
• LLM-based analysis revealed how therapist language can guide clients toward psychological distancing.
• The LLM-powered measure reliably correlated with patient symptom severity and treatment progress.

Conclusions:

• LLMs significantly enhance the ability to model the relationship between linguistic distance and psychopathology.
• This approach expands the analysis to include therapist language, offering deeper insights into therapeutic mechanisms.
• LLM-powered linguistic analysis holds potential for advancing the detection and treatment of psychopathology.