Title: Establishing Meta-Decision-Making for AI: An Ontology of Relevance, Representation and Reasoning

Abstract: We propose an ontology of building decision-making systems, with the aim of establishing Meta-Decision-Making for Artificial Intelligence (AI), improving autonomy, and creating a framework to build metrics and benchmarks upon. To this end, we propose the three parts of Relevance, Representation, and Reasoning, and discuss their value in ensuring safety and mitigating risk in the context of third wave cognitive systems. Our nomenclature reflects the literature on decision-making, and our ontology allows researchers that adopt it to frame their work in relation to one or more of these parts.