Leveraging User-Centered and Neuro-Symbolic Artificial Intelligence in Information Systems

Modern Artificial Intelligence (AI) has emerged as a transformative force within Information Systems (IS), playing an increasingly critical role in modern organizations. This importance is evidenced not only by substantial investments but also by its widespread presence in mainstream applications, such as AI-driven chatbots and AI-enhanced search technologies. As digitalization efforts continue, the advent of cutting-edge technologies, particularly through Large Language Models (LLMs), necessitates ongoing organizational adaptation to maintain competitiveness. These technologies offer considerable opportunities, including enhanced decision-making, improved risk and knowledge management, better strategic alignment, and optimized supply chains with respect to both financial outcomes and sustainability. Despite these opportunities, current developments resemble the early, unregulated phase of the World Wide Web, commonly referred to as the “wild west”, and are accompanied by substantial disruptions and risks. Established user interaction paradigms are increasingly obsolete, traditional modes of knowledge consumption are shifting, and foundational IS processes must be re-envisioned. In response, this thesis advocates for the integration of user-centered and neuro-symbolic AI approaches that harness AI capabilities while anchoring human expertise at the core of system design. Central to this vision is the use of knowledge-based techniques, particularly Knowledge Graphs (KGs), which semantically interlink information and, when combined with LLMs, substantially enhance knowledge management and organizational insight. This thesis introduces two novel, user-centered and neuro-symbolic AI approaches that enable the transformation of structured (relational databases) and unstructured (text) organizational data into semantically rich KGs. Given that such data are often underutilized and poorly connected, these approaches facilitate both human and machine interpretability. They support automation, increase efficiency, and provide confidence scores to foster user trust and usability. Building on these KGs, three AI-driven procurement approaches are proposed to advance data analytics, automation, sustainable sourcing, and information handling—areas typically constrained by manual, time-intensive processes. These include: (1) a digital procurement workspace supporting process automation and analytics, (2) a customizable dashboard framework tailored to diverse user expertise levels, and (3) a system for automated, concurrent negotiations, particularly vital in contexts marked by geopolitical (tariffs and sanctions) or pandemic-inducedsupply chain disruptions. All proposed solutions prioritize user acceptance, usability, and experiential quality while unlocking new potential for efficiency and cost reduction. Finally, a novel goal-oriented recommendation system is introduced, addressing limitations in traditional recommendation systems that struggle with intent inference. By explicitly capturing user goals, this approach enables stepwise, goal-aligned product recommendations, thereby enhancing the relevance, effectiveness, and user satisfaction of recommendations. Collectively, this thesis contributes a comprehensive collection for embedding user-centered and neuro-symbolic AI into modern IS, demonstrating tangible benefits in data transformation, procurement, and intelligent recommendation systems.