Landscape of AI safety concerns -- A methodology to support safety assurance for AI-based autonomous systems
Ronald Schnitzer, Lennart Kilian, Simon Roessner, Konstantinos Theodorou, Sonja Zillner
TL;DR
The paper tackles the challenge of assuring safety for AI-based autonomous systems by addressing the AI-specific safety concerns that create a semantic gap between intended and observed behavior. It introduces the Landscape of AI Safety Concerns (LAISC), a methodology that enumerates AI-SCs, decomposes them into concrete, verifiable goals, derives verifiable requirements, and applies Metrics and Mitigation Measures along the AI life cycle to assemble evidence for safety assurance cases. The core contributions are a structured, repeatable process to demonstrate the absence of AI-SCs through VR-driven evidence, and a concrete case study on a driverless regional train that illustrates practical instantiation, tabular reporting, and argumentation support. The approach complements traditional safety analyses and standards, offering a rigorous framework to strengthen confidence in AI safety while acknowledging that it is one component of a holistic safety assurance strategy.
Abstract
Artificial Intelligence (AI) has emerged as a key technology, driving advancements across a range of applications. Its integration into modern autonomous systems requires assuring safety. However, the challenge of assuring safety in systems that incorporate AI components is substantial. The lack of concrete specifications, and also the complexity of both the operational environment and the system itself, leads to various aspects of uncertain behavior and complicates the derivation of convincing evidence for system safety. Nonetheless, scholars proposed to thoroughly analyze and mitigate AI-specific insufficiencies, so-called AI safety concerns, which yields essential evidence supporting a convincing assurance case. In this paper, we build upon this idea and propose the so-called Landscape of AI Safety Concerns, a novel methodology designed to support the creation of safety assurance cases for AI-based systems by systematically demonstrating the absence of AI safety concerns. The methodology's application is illustrated through a case study involving a driverless regional train, demonstrating its practicality and effectiveness.