Leveraging natural language processing to study emotional coherence in psychotherapy

Area of Science:

• Psychology
• Computational Linguistics
• Psychotherapy Research

Background:

• Emotional coherence, the alignment between emotional experience and expression, is crucial for psychological functioning.
• Natural Language Processing (NLP) advancements enable automated emotion recognition in therapeutic dialogues, allowing for large-scale analysis of emotional coherence.
• Previous research on emotional coherence was limited by sample size and analytical granularity.

Purpose of the Study:

• To automatically label client emotions using NLP models in psychotherapy dialogues.
• To investigate the coherence between verbally expressed and self-reported emotions.
• To examine the association between emotional coherence and improvements in client functioning during therapy.

Main Methods:

• Utilized state-of-the-art emotion recognition models (transformer-based) to label 139,061 utterances from 872 transcribed therapy sessions.
• Collected client self-reported functioning and emotions data.
• Employed multilevel modeling to assess emotional coherence and its link to functional changes.

Main Results:

• The NLP emotion recognition model achieved moderate performance.
• A significant association was found between clients' verbally expressed and self-reported emotions.
• Greater coherence in negative emotions correlated with improved client functioning throughout treatment.

Conclusions:

• A link exists between clients' subjective emotional experiences and their verbal expression, highlighting the importance of this emotional coherence for functioning.
• NLP tools can effectively analyze emotional processes in psychotherapy, offering new insights into therapeutic outcomes.
• Coherence in negative emotional expression appears particularly important for positive therapeutic change.