The ECCA Model: Adaptive Knowledge Modeling to Support Human-AI Collaboration in Dynamic Contexts

Abstract

Safety-critical domains such as oil and gas or aviation operate as complex socio-technical systems characterized by continuous change. In such contexts, knowledge is often modeled using static representations that fail to capture temporal, contextual, and tacit aspects of real-world work. This disconnect becomes particularly problematic as artificial intelligence (AI) systems are increasingly introduced to support operations. Many AI models rely on fixed worldviews, limiting their adaptability, while symbolic approaches often lack the flexibility required for high-stakes environments. This study investigates how to model domain knowledge in ways that are semantically structured yet dynamically adaptive. Through a structured literature review and synthesis of recent research, we identify key limitations of static models and explore conceptual and technical strategies to overcome them. We then propose the ECCA model (Elicitation, Contextualization, Comparison, and Adaptation) as a multi-layered framework to support human-AI collaboration in dynamic environments. The model integrates ontological representations for semantic consistency, neuro-symbolic AI for learning and adaptation, and participatory mechanisms to incorporate tacit expertise. Two hypothetical scenarios, in offshore drilling and commercial aviation, demonstrate the practical application of the model, showcasing its capacity to evolve domain knowledge while ensuring traceability and alignment among stakeholders. ECCA contributes to bridging formal modeling techniques with real-world operational needs by enabling knowledge systems that evolve alongside the domains they represent. Future work will explore empirical validation and integration into engineering workflows across safety-critical sectors.