Untangling Shared Phenomena for Improving Analysis of Normative Requirements
Junwei Quan, Lina Marsso, Dalal Alrajeh, Marsha Chećhik · 2025
Systems that interact with humans, such as assistive robots, must adhere to normative requirements to avoid social, legal, ethical, empathetic, or cultural (SLEEC) harms. However, these requirements are inherently complex, involving interdisciplinary input and time-sensitive constraints. This complexity poses challenges for stakeholders in accurately specifying and ensuring the system’s ability to meet these requirements without conflicts. Recent research introduced a domain-specific language and validation tools to help stakeholders express SLEEC rules, emphasizing consistency and reducing redundancy. However, these tools often overlook whether the prescribed behaviors are realizable given the system’s capabilities and operational environment. This oversight can lead to requirements that depend on environmental factors or user actions beyond the system’s control, rendering some rules unrealizable. This paper addresses this gap by analyzing normative requirements through the lens of system and environmental capabilities across nine use cases. We apply controller synthesis techniques to assess whether the system can fulfill these requirements based on its defined capabilities. This approach helps identify faults, resolve conflicts, and provide actionable feedback to support interdisciplinary stakeholders to improve quality of requirements.