Exploring data augmentation and memory strategies for AI-based synthetic personae

In this thesis, we investigate the enhancement of synthetic personae through data augmentation and memory strategies, focusing on a case study involving the development of a chatbot impersonating Van Gogh. Leveraging large language models (LLMs), we explore the integration of a novel autonoetic memory dataset derived from Vincent Van Gogh's biography and letters to improve the chatbot's question-answering capabilities by accessing different layers of memory. This research not only delves into the potential of LLMs for creating engaging synthetic personae but also addresses the challenges of data augmentation and the practical implementation of memory systems in chatbots. Through comparative analysis, we demonstrate the superiority of the proposed approach over traditional models, highlighting its contributions to the fields of Human-Computer Interaction (HCI) and synthetic personae development. This work sets the stage for future exploration in enhancing chatbot interactions and opens new avenues for research in cognitive model design.