토픽 모델링을 통한 심리상담 챗봇의 개입 유형별 사용자 경험 분석

초록

The recent advancements in generative AI has led to the widespread adoption of mental health chatbots that enhance individuals' access to self-help psychological support. Despite the growing academic interest, however, efforts to examine user experiences across different intervention types remain limited. This study examined the user experiences between two types of interventions: companionship-based and CBT-based chatbots. We collected approximately 130,000 English-language reviews from two representative apps―Replika and Wysa―and applied Latent Dirichlet Allocation (LDA) to extract key user experience topics. These topics were then mapped onto five overarching themes for structured comparison. The findings revealed that interactivity-related themes were predominant in Replika reviews, whereas themes related to psychological support and CBT-specific interventions were more salient in Wysa. Through a large-scale analysis of user-generated text, this paper advances understanding of chatbot user experiences across intervention types and offers insights into tailored design strategies and data privacy policies.

키워드