Integration of a Chatbot to aid Young Carers

Abstract

This study explores the integration of a chatbot designed to assist young carers, focusing on developing, integrating, and evaluating a proof-of-concept (PoC) to address their unique needs. The research is guided by three questions: the application and tuning of multinomial Naive Bayes (NB) for training the chatbot, the technological and ethical challenges encountered in the development process as informed by an evaluation of related work, and the identification of critical features in advanced language models, like Large Language Models (LLMs), that contribute to creating an effective support tool for young carers. A mixed-method approach was employed, combining the chatbot development using the React framework and a multinomial NB algorithm, with a subsequent analysis indicating the potential benefits of transitioning to LLMs for enhanced user engagement. The study also contrasts the expert-driven methodology utilized in the chatbot’s development with the participatory approaches found in existing research, highlighting the implications for user engagement and support efficacy. Significant findings include the practicality of the chosen technologies and methodologies in the initial development phase, the potential for LLMs to improve the chatbot’s responsiveness and interaction quality, and the identification of both technological and ethical challenges in chatbot implementation. The thesis concludes with recommendations for future improvements, emphasizing the integration of advanced natural language processing techniques and ethical considerations in developing chatbots for sensitive populations like young people.

Date of Award	2024
Original language	English (American)
Supervisor	Marc Kurz (Supervisor) & Alexander Palmanshofer (Supervisor)