Situational decision making for autonomous systems with SLEEC requirements

Robotic and autonomous systems are increasingly deployed in settings where complex human norms play an important role. Ensuring these systems conform to Social, Legal, Ethical, Empathetic, and Cultural (SLEEC) norms is extremely challenging. In particular, traditional design-time evaluation of conformance — which considers all possible scenarios, including worst cases — often yields unnecessarily strong constraints. This rigidity impedes adaptation to the specific situations actually encountered in operation, where not all theoretical worst-case scenarios materialise. Effective adaptation requires situational decision making to manage trade-offs in the presence of possibly conflicting requirements, and principled methods for prioritising one norm over another if and when necessary. Our paper addresses this need by providing a formal semantic foundation for situational SLEEC conflict resolution. We extend the timed process-algebraic semantics of a recently introduced SLEEC domain-specific language with support for rule prioritisation. We define two novel mechanisms (cancellation-based and conflict-aware prioritisation) and formalise them using a tock-CSP prioritisation operator. This supports both strict overrides and “lazy” conflict resolution. We demonstrate the utility of our formal framework and its support for mechanical reasoning by using the FDR model checker to verify illustrative SLEEC rules for an assistive-robotics scenario.