Impact of Artificial Intelligence in the Certification of Human-Centered Aviation systems

Abstract

In recent years we have witnessed the emergence of applications based on artificial intelligence in the aviation industry. This technology is said to be promoting a new era or evolution, such as the introduction of jet engines in the 1950s and fly-by-wire in the 1980s. To maintain aviation safety standards in this transition, civil aviation authorities responsible for certifying aerospace systems must anticipate the unprecedented impact of AI on human-centric aerospace systems and answer a number of critical questions:•How to establish public trust in human- centric AI-based systems?•How to integrate the ethical dimension of human- centric AI (transparency, non-discrimination, fairness, etc.) in safety certification processes? •How to prepare for the certification of human- centric AI systems? •What standards, protocols, methods needs to be developed to ensure that human- centric AI further improves the current level of air transport safety? EASA, the European Aviation Safety Agency, has recently developed a roadmap for the certification of AI applications in aviation, which analyzes the involvement of human- centric AI in the aviation sector and identifies the objectives that must be met and the actions that must be taken to respond to the previous questions. This effort constitutes a starting point for the certification of human- centric AI in aerospace systems. It develops in particular the core notion of trustworthiness of human- centric AI in human centered systems and proposes a framework based on four human- centric AI trustworthiness building block:— trustworthiness analysis — learning assurance — explainability — safety risk mitigation The presented paper syntheses the concept of human- centric AI applications, it also discusses and revises the 4 elements of the trustworthiness of human- centric AI framework proposed by EASA, and based on this discussion anticipates the possible impacts of the introduction of human- centric AI in the different Implementation Rules (IR), Certification Specifications (CS), Acceptable Means of Compliance (AMC) and guidance material (GM) in the domains covered by the EASA Basic Regulation.

Keywords: Human Centric, Aviation, AI, Certification

DOI: 10.54941/ahfe1001432