A Deontic Logic Analysis of Autonomous Systems' Safety

Abstract: We consider the pressing question of how to model, verify, and ensure that autonomous systems meet certain \textit{obligations} (like the obligation to respect traffic laws), and refrain from impermissible behavior (like recklessly changing lanes). Temporal logics are heavily used in autonomous system design; however, as we illustrate here, temporal (alethic) logics alone are inappropriate for reasoning about obligations of autonomous systems. This paper proposes the use of Dominance Act Utilitarianism (DAU), a deontic logic of agency, to encode and reason about obligations of autonomous systems. We use DAU to analyze Intel's Responsibility-Sensitive Safety (RSS) proposal as a real-world case study. We demonstrate that DAU can express well-posed RSS rules, formally derive undesirable consequences of these rules, illustrate how DAU could help design systems that have specific obligations, and how to model-check DAU obligations.