Counselor-AI Collaborative Transcription and Editing System for Child Counseling Analysis

DOI

Psychological counseling, especially for children, heavily relies on capturing both verbal and non-verbal cues to understand and support each child's emotional and developmental needs. Therefore, creating a detailed and accurate transcription for a child’s counseling session is crucial but often labor-intensive and time-consuming, which makes it challenging to maintain the consistency of counseling quality. Despite advancements in AI, current session analysis practices rely primarily on manual clinical assessments and struggle to accurately capture children's verbal and non-verbal expressions. To address these challenges, we propose an AI-based expert support system designed to enhance child counseling analysis. The system comprises two key components: (i) a transcription generation model and (ii) an editable dashboard. The transcription generation model extracts verbal expressions from both children and counselors, verifies speakers' identities, and objectively captures non-verbal cues using a Multimodal Large Language Model. The editable dashboard facilitates Counselor \& AI collaboration, where AI reduces human bias by providing objectivity, and counselors mitigate the risk of over-reliance on AI while maintaining oversight. This collaboration ultimately enhances workflow efficiency and leads to accurate counseling analyses. An evaluation with 48 child counselors demonstrates the system's superior effectiveness and usability compared to existing services, with a majority expressing a strong intent to continue using our system. The system not only improves transcription accuracy but also supports more precise analysis of counseling sessions, enabling counselors to focus more on therapeutic engagements. These findings highlight the system's potential to reduce the workload of child counselors, improve the quality of counseling services, and provide valuable resources for both individual counseling and counselor training. To the best of our knowledge, our study is the first to propose an AI-based expert support system optimized for generating transcriptions for child counseling analysis.