AI-enhanced therapeutic alliance: bridging AI innovation and mental health care

Purpose The current setback associated with traditional psychotherapy is its limited accessibility and time constraints. To overcome these limitations, this study aims to propose an artificial intelligence (AI)-powered platform methodology for psychotherapy, which offers depression severity evaluation, journaling support, real-time therapist recommendations and automated conversation summaries. Design/methodology/approach The platform incorporates Bidirectional Encoder Representations from Transformers–based classifiers for multi-label classification of depression causing emotions, scoring of depression severity and detection of emotions from patient journals. Along with that, Bidirectional and Auto-Regressive Transformers–based transformer model is used for producing short summaries of patient-therapist conversation. The classifier models were finetuned on tagged transcripts. Findings The model showed an accuracy of 96% in Emotion classification and 98% in Sentiment Classification, having high precision, recall, F1-Score and Support. Also text-generation in Patient Interaction Module and Summary Generation part with good Recall-Oriented Understudy for Gisting Evaluation scores. The approach of sentiment classification and semantic similarity showed its potential usage in depression detection and emotion tracking. And also showcased its potential competition to traditional approach of regression-based severity measurement. Originality/value By integrating AI methods into a single therapeutic workflow, the platform maximizes therapist efficiency and reduces case-loads with increasing therapeutic needs. Also, the inclusion of journaling by patients, which is a coping mechanism prescribed by therapist and analysis of it provides insights of patient’s recovery to the therapist in frequent manner.