NLP meets psychotherapy: Using predicted client emotions and self-reported client emotions to measure emotional coherence

Abstract: Emotions are experienced and expressed through various response systems. Coherence between emotional experience and emotional expression is considered important to clients' well being. To date, emotional coherence (EC) has been studied at a single time point using lab-based tasks with relatively small datasets. No study has examined EC between the subjective experience of emotions and emotion expression in therapy or whether this coherence is associated with clients' well being. Natural language Processing (NLP) approaches have been applied to identify emotions from psychotherapy dialogue, which can be implemented to study emotional processes on a larger scale. However, these methods have yet to be used to study coherence between emotional experience and emotional expression over the course of therapy and whether it relates to clients' well-being. This work presents an end-to-end approach where we use emotion predictions from our transformer based emotion recognition model to study emotional coherence and its diagnostic potential in psychotherapy research. We first employ our transformer based approach on a Hebrew psychotherapy dataset to automatically label clients' emotions at utterance level in psychotherapy dialogues. We subsequently investigate the emotional coherence between clients' self-reported emotional states and our model-based emotion predictions. We also examine the association between emotional coherence and clients' well being. Our findings indicate a significant correlation between clients' self-reported emotions and positive and negative emotions expressed verbally during psychotherapy sessions. Coherence in positive emotions was also highly correlated with clients well-being. These results illustrate how NLP can be applied to identify important emotional processes in psychotherapy to improve diagnosis and treatment for clients suffering from mental-health problems.

• The paper demonstrates that NLP techniques can accurately correlate self-reported emotions with verbal expressions in authentic psychotherapy sessions.
• The paper employs AlephBERT with rigorous 10-fold cross-validation on Hebrew datasets to enhance emotion recognition performance.
• The paper finds a strong positive link between emotional coherence in positive emotions and client well-being, suggesting valuable diagnostic applications.

Emotional Coherence in Psychotherapy through NLP

The paper "NLP meets psychotherapy: Using predicted client emotions and self-reported client emotions to measure emotional coherence" (2211.12512) presents a compelling approach to understanding emotional coherence within psychotherapy sessions. By leveraging NLP techniques and emotion recognition models, this research investigates the coordination between emotional experiences and expressions in therapy. Furthermore, it assesses how this coherence correlates with client well-being across sessions.

Introduction and Background

Emotional coherence, defined as the relationship between emotional experiences and their expressive counterparts, plays an integral role in therapeutic practices. However, past research primarily conducted in controlled environments lacks the temporal depth and authenticity present in real therapy sessions. Existing methodologies often hinge on isolated evaluative measures that fail to capture evolving emotional responses over time. The current study fills this gap by employing NLP to analyze psychotherapy interactions, offering a scalable approach that transcends the limitations of previous experiments dependent solely on human annotators.

Methodological Framework

The researchers embark on testing two primary hypotheses: first, that there is a positive correlation between verbal expressions of emotion and clients' self-reported emotional experiences; and second, that enhanced emotional coherence is associated with improved client well-being. Utilizing a Hebrew psychotherapy dataset (BIU-872), which contains both annotated (BIU-872_Gold) and unannotated (BIU-872_Silver) sessions, the methodology employs emotion annotations at the utterance level for clients using AlephBERT, a transformer-based model tailored for Hebrew text classification tasks.

Model Implementation

AlephBERT, adapted for Emotion Recognition (ER), enables the annotation of emotional data by training on the BIU-872_Gold dataset through a rigorous 10-fold cross-validation process. It addresses domain-specific challenges by utilizing pre-trained LLMs initially calibrated on substantial Hebrew language corpora. Key hyperparameters such as learning rate and token size are fine-tuned to optimize performance. As a result, the pre-trained model labels client emotions in BIU-872_Silver, facilitating comprehensive analysis across extensive datasets.

Key Findings

Emotional Coherence Validation

The research identifies a statistically significant positive correlation between self-reported positive (P_pos) as well as negative emotions (P_neg) and their verbal expressions (U_pos, U_neg). This coherence is validated using traditional and automated NLP methods, marking the first instance of a large-scale empirical validation that extends beyond controlled lab settings to authentic therapeutic environments. The application of NLP models ensures scalability and allows for a more nuanced exploration of emotional processes unseen in prior studies.

Well-being Correlation

The study further reveals that coherence between positive emotional experiences and expressions strongly correlates with higher scores in well-being assessments. Notably, the positive association is absent in negative emotional coherence, implicating that effective therapeutic outcomes are more closely aligned with positive emotional processes. The results indicate that the NLP-based model could serve as a diagnostic tool, accurately detecting emotional coherence and informing therapeutic interventions to enhance client well-being.

Conclusion

The integration of NLP and psychotherapy offers a powerful framework for understanding emotional dynamics in therapy. By using advanced emotion recognition techniques, this paper effectively bridges the gap between subjective emotional experiences and objective verbal expressions over the course of therapy. These insights contribute to both the theoretical and practical applications in psychotherapy, enabling personalized treatments tailored to individual emotional profiles. Future research may further explore cross-linguistic applications and refine diagnostic models to undertake a broader spectrum of emotional and psychological evaluations within therapeutic settings.