Augmenting Knowledge Graph-Based Decision Support Systems For Smes By User Collaboration

Data- and AI-based Decision Support Systems (DSS) offer huge potential for industrial applications. Despite increasing levels of automation, many AI systems suffer from a lack of data, which leads to the need for user intervention and interaction (collaboration). This is particularly true for AI systems that incorporate knowledge graphs (KGs) for link prediction using knowledge graph embeddings, as KGs are inherently sparse by nature. This paper presents a real-world use case in a medium-sized enterprise in which users collaboratively interact with a KG-based DSS. In this setting, the employees are assisted by AI in decision-making while simultaneously improving the AI by providing new data. We contribute by (1) showing that small improvements in data can contribute to significant improvements of AI output and (2) providing steps to ensure the translation from technical AI components to a DSS suited for human-AI collaboration. The insights are generalized into practical guidance for small and medium-sized enterprises that lack mature data management capabilities, outlining how to approach AI adoption in a human-centered and collaborative manner.