An ontology-guided knowledge model for requirements engineering of trustworthy AI systems

• Proposes OG-RE4AI: A conceptual Knowledge Model for trustworthy AI systems • Integrates Domain, Inference, and Task layers to operationalize RE goals • Domain and Inference layers constitute an ontology-guided knowledge model • Questionnaire operationalizes logic into a guided, goal oriented RE process • Validated via SLR and interviews with 20 safety-critical practitioners Achieving a trustworthy system is a fundamental goal for all stakeholders involved in system development. Developing trustworthy systems requires balancing functional goals with complex socio-technical constraints. However, the probabilistic and data-driven nature of Artificial Intelligence (AI) introduces abstract demands such as explainability and human oversight, that are difficult to operationalize, creating specification "blind spots." Traditional Requirements Engineering (RE) methods often lack the structured knowledge representation needed to systematically manage this complexity. To address this challenge, we propose an Ontology-Guided Requirement Engineering for AI (OG-RE4AI), a comprehensive knowledge-based framework. At its core lies a unified knowledge model that integrates three functional layers: (1) Ontology-guided conceptual schema that defines the structures and the semantic organization of AI-powered systems requirements; (2) an instantiated knowledge base (operationalized via a structured questionnaire) that translates abstract requirements into executable RE lifecycle items; and (3) a decision-support process that utilizes these inputs to facilitate the systematic negotiation of requirement trade-offs. The framework was developed through a Systematic Literature Review (SLR) and validated via semi-structured interviews with 20 practitioners from safety-critical and high-assurance sectors. By providing a comprehensive representation of AI requirements while explicitly exposing their latent interdependencies, OG-RE4AI facilitates transparent stakeholder collaboration and informed decision-making. This holistic visibility enables development teams to effectively align technical, ethical, and organizational priorities, ensuring the rigorous governance of trustworthy AI-powered systems.