Virtual companionship : Intelligent custom companionship software

Abstract

Loneliness has been identified as a predictor of depression in people, especially in adolescents, with communication recognized as a potential means to alleviate feelings of loneliness. The ad-vent of large language model (LLM) chat systems has enabled the development of robots capable of engag-ing in conversations with individuals at any time and place. This study presents the creation of a chatbot system designed to dynamically adjust its conversational style based on the 16 personality types of the My-ers-Briggs Type Indicator (MBTI). By utilizing fine-tuning techniques for large language models, the system adapts its responses and tone to align with the selected personality type, offering users a personalized and engaging interaction. The system was implemented using Python Flask for the backend, incorporating fine-tuned language models to facilitate natural language processing, and the frontend Android application was developed in Java to provide a user-friendly interface. The evaluation results indicate that the chatbot effectively mirrors per-sonality-specific traits within conversations.