Development and Evaluation of AI Chatbot Personas for Event Planning: A Comparative Study of Casual and Professional Interaction Styles

Modern AI tools have made the process of developing chatbots easier than ever. This thesis explores designing and developing two chatbot personas for an event planning and networking application as well as a study about the user interactions with the personas in terms of conversational clarity, relevance of information, and user engagement. The project began with the design of the two unique personas through a psychology framework called the Five Factor Model. Next, the design and architecture of the chatbot was created, using a loose retrieval augmented generation (RAG) structure, and a two layer API call to OpenAI’s API models. The main layer selects an intent function which is responsible for fetching relevant information from the database, and the second uses the returned data to formulate an appropriate response to the user. This backend server had a single endpoint. The frontend was created using React.js. To conduct the study, users were asked to interact with both chatbots, while reflecting on their interactions with both. Findings from the study indicate that there was no significant difference in users’ perception of conversational clarity or relevance of information between the two chatbot personas which suggests that both were effective in delivering accurate and understandable content. However, there was a noticeable and statistically significant increase in user engagement with the casual chatbot. The results indicate that while clarity and relevance of information may not depend heavily on the persona, engagement is influenced slightly by chatbot persona.