사용자 인식 기반 챗봇 지능의 다차원 구조 탐색: 레퍼토리 그리드(Repertory Grid) 기법을 활용하여

초록

This study aims to explore the components and multidimensional structure of perceived chatbot intelligence from the user's perspective. Unlike prior research that has primarily defined AI intelligence based on technical performance metrics, this study focuses on how users form perceptions of intelligence through actual interaction experiences. To achieve this, the Repertory Grid Technique (RGT), grounded in Personal Construct Theory, was employed to elicit the constructs users rely on when judging chatbot intelligence, and ratings for each construct were collected across different chatbots. Content analysis and hierarchical cluster analysis revealed that perceived chatbot intelligence is not a single dimension but a multidimensional structure. It was structured into five key dimensions: System Efficiency, Logical Expertise, Contextual Reliability, Practical Utility, and Social-Emotional Intelligence. These findings suggest that chatbot intelligence is evaluated not only in terms of information processing but also through cognitive, functional, and social interactional qualities. This study contributes theoretically by extending the concept of AI intelligence from a technology-centered view to a user perception-centered perspective. Methodologically, it demonstrates the applicability of the Repertory Grid Technique in AI research by uncovering users' cognitive structures in a bottom-up manner. The findings also offer practical implications, suggesting that chatbot design should emphasize contextual understanding, explainability, emotional responsiveness, and information credibility.

키워드