Human-Centered Conversational AI Design Artifacts Leveraging Semantic Techniques AND Large Language Models: Meta Requirements and Design Principles and Design Patterns

Conversational agents (CAs) have emerged as powerful tools for human-computer interaction, offering natural language interfaces and personalized assistance, particularly in healthcare. The development of CAs leveraging knowledge engineering (KE) techniques and LLMs necessitates the consideration of specific design principles (DPs) and interactive patterns to ensure their effectiveness and user satisfaction. This research focuses on a comprehensive analysis of the DPs and patterns crucial for developing CA artifacts using KE techniques. To address the gap, author incorporated design science research methodology (DSRM) principles tailored to the modelling workshop method to capture domain knowledge and practitioner expertise in a new health information system class. The research explores DPs based on meta-requirements suitable for interaction patterns and KE-based CAs, including ontology-driven approaches. The research proposed systematic pathways, and by applying these DPs and patterns, developers can develop CAs exploiting LLMs to understand complex user queries, provide accurate responses, and adapt to dynamic contexts.