Agentic LLM Workflows for Personalized User Experience Questionnaire Generation
Yeonwoo Kim, Junhyeok Lee, Jaehyun Han, Minjae J. Kim, Howook Lee, Won Hee Lee · 2024
Effective user experience (UX) evaluation requires personalized assessment methods that adapt to individual user characteristics and real-time context. This study introduces the User Experience Questionnaire generation workflow using multiple LLMs (UEQ-mLLM), a system that generates tailored questionnaires based on user data collected through a multimodal interactive dashboard. By leveraging user information and states, UEQ-mLLM generates questionnaires that enhance the accuracy and depth of UX evaluations. Comparative analysis against a single LLM-based approach using the G-Eval framework demonstrated a significant performance improvement of 20.62% for UEQ-mLLM. This work highlights the potential of utilizing multiple LLMs to generate effective UX questionnaires and contributes to the advancement of user-centered design methodologies.