Presentation Information
[4Yin-B-46]A Motivational Interviewing Dialogue System with Summary Utterance Prediction: Development and Evaluation
Hiroki Tsuchiya1, 〇Jie Zeng1, Fumio Nihei1, Yukiko Nakano1 (1. Seikei University)
Keywords:
Dialogue systems,Motivational interviewing,Embodied conversational agents
This study develops and evaluates a Motivational Interviewing (MI) dialogue system that predicts appropriate timing for summary utterances and generates context-aware summaries. The system integrates three modules: a basic MI-consistent response generator, a summary generator that organizes dialogue content, and a prediction module that selects the appropriate response type. It dynamically switches responses based on dialogue context and incorporates speech I/O and CG animation to create a multimodal counseling environment. Experiments comparing the system with baselines lacking summary functionality showed that the summary module improved summary quality while maintaining MI adherence. However, it did not sufficiently enhance change talk frequency and faced issues regarding the natural timing of summaries. These results indicate partial effectiveness of the proposed system but highlight the need for further improvement of the summary function.