The multi-entity decision graph decision ontology: A decision ontology for fusion support

Abstract

Aiding decision-makers is a key function of a fusion system. In designing decision-aiding modules for fusion systems, it is necessary to understand the elements of the decision model and the dependencies that connect them. An ontology is a disciplined means to codify that understanding. Many fusion systems have a Bayesian Network (BN) component to support probabilistic reasoning under uncertainty. Decision graphs (DG) are an extension that adds decision aiding to BNs. Both BNs and DGs have limited logical expressivity, able to model propositions, but cannot directly model variable numbers of entities or variations in their attributes and relationships. This important capability is called first-order expressivity. Multi-Entity Bayesian Network (MEBN) was developed to provide first-order logic expressivity to BNs. We are developing Multi-Entity Decision Graph (MEDG) to do the same for decision graphs. We found that a decision ontology is useful to our efforts. The literature has a limited discussion of decision ontologies. Almost all focus on the entities and the entity hierarchy. But BNs and DGs emphasize relationships and the dependencies between relationships. The key for probabilistic first-order expressivity is to identify the relationships that enable dependencies between entity instances. We developed a MEDG Decision Ontology that highlights both the entities and key relationships that any decision model needs to address. It is designed to support decision model developers, including fusion model developers, in building comprehensive decision aiding capabilities.