Customizable AI for Depression Care: Improving the User Experience of Large Language Model-Driven Chatbots

Large Language Models (LLMs) demonstrate significant potential in the field of mental health. However, existing chatbots often lack personalized designs, which may limit their ability to fully address the complex needs of users with depression. This study builds upon the previously developed CloudEcho system, a mental health management application that integrates emotion monitoring and psychological support functions, to explore the impact of role customization features on user trust and customization experience. Using a mixed-methods approach, this study compares the differences between the system with role customization features and its original version. Quantitative results indicate that role customization can enhance user trust, showcasing high usability and satisfaction. Qualitative interviews further reveal the strengths and limitations of this feature and suggest directions for optimization. Together, these findings highlight the potential value of chatbot role customization in mental health support and offer theoretical and practical guidance for future LLM-driven personalized design and optimization in mental health contexts.