Design Features for Explainable Generative AI (GenXAI) Systems in Knowledge-Intensive Service Work

Abstract

The use of generative AI (GenAI) and large language models (LLMs) in knowledge-intensive fields like customer support is rapidly growing. While GenAI responses often appear persuasive, they carry the risk of inaccuracies and hallucinations. Hence, users must critically evaluate responses to reach appropriate reliance and knowledge utilization. Despite technological advancements, design knowledge for enhancing human-GenAI interaction from an explainable AI (XAI) perspective remains lacking. Thus, this study applies the design science research (DSR) approach to develop explanations that aid human interaction with GenAI systems. Drawing from XAI literature and human reasoning theories, we built and evaluated seven design features and instantiated a prototype that contributes to the development of reliable explainable GenAI (GenXAI).

Ähnliche Arbeiten