Trust Evaluation of Ontological Decision Support Systems for Avionics Analytics

Artificial Intelligence (AI), including Machine Learning (ML), has gained prominence to perform data analytics in different application domains such as medicine, government, and avionics. The main challenge for AI/ML-based systems adoption in avionics requires trust the technology to support or carry out given tasks, e.g., decision making. This paper discusses a method to assess trustability in an ontological Decision Support System (DSS) for avionics analytics in Air Traffic Management (ATM). The approach leverages the Multisource AI Scorecard Table (MAST) which provides aviation stakeholders (e.g., developers and users) with a standard checklist. The MAST checklist is aligned with the principles of good analysis adopted by the Intelligence Community (IC) which helps promote the development of more understandable system to engender trust in AI outputs. Such a scorecard provides information-guided assessment of transparent and consistent understanding of AI/ML tools applied for commercial and government aviation use. This paper explores the application of the MAST to a DSS driven by an Avionics Analytics Ontology (AAO). It shows how a AAO-based DSS can be assessed (as to trustability) by means of a case study with two application scenarios set by different airspace situations for autonomy and collision avoidance.