From Human Hands to Artificial Intelligence

Abstract

Artificial Intelligence (AI) is often regarded as a technology that can revolutionize the relationship between humans and machines. Concerns regarding these potential novel interactions have motivated the drafting of regulatory and industry guidance specifically for systems that are enabled with AI. This perception of human-like interaction, such as the conversational interaction with ChatGPT, can lead users and policymakers to believe that AI is challenging how we should think about these systems entirely, though experience in systems design and aircraft type certification shows this is not truly the case. Misleading, confusing, and opaque system designs have been key concerns for aircraft designers and regulators for decades. Numerous civil aviation accidents and incidents have been the motivation for the development of regulations and policies regarding Human Factors (HF) in the civil aviation domain. The HF discipline takes a distinct perspective of the aircraft in comparison to regulations and policy regarding systems design and behavior. For this paper, a perspective of Human Factors is applied which aligns with 14 CFR 25.1302. This regulation applies to “installed systems and equipment intended for flightcrew members' use in operating the airplane from their normally seated positions on the flight deck.” This perspective is compared to complementary regulations of general applicability which drive the design of aircraft systems and equipment such as 14 CFR 25.1301 and 14 CFR 25.1309. These regulations and their associated policy and guidance provide a framework for understanding the distinct needs of the human and machine versus the AI that may enable it. This paper investigates this framework, and the System Engineering infrastructure it relies on, to show these perspectives are agnostic to the technology used to develop or implement systems on aircraft. It also highlights that there is a tangible linkage between them which can be used to better inform both perspectives. By leveraging this framework and existing processes, this paper will show that the current separation of perspectives is well-suited for many forms of AI implementations. Further, it discusses enhancements to the linkage between these perspectives that can be leveraged to address the expectations of the operator as well as the needs of the development enabled by AI. These enhancements reinforce the framework with metrics that allow the flow of development information up and down the hierarchy of design detail.

Ähnliche Arbeiten