Human-Robot Ethical Collaboration Through Adaptive Business Processes

Robots are increasingly deployed in human environments to collaborate and interact with people. This growing integration has raised ethical concerns that extend beyond compliance with “hard” normative requirements to include subjective moral values and preferences. Consequently, there is a need to design systems that preserve both regulations and subjective values. Business processes are commonly used to specify how systems respond to events, coordinate with users, and achieve their objectives. More recently, they have been adopted to define the behavior of robotic systems, including extensions that incorporate ethical constraints. However, static, one-size-fits-all workflows lack the flexibility to adapt to dynamic runtime contexts, diverse user needs, and evolving ethical preferences. The challenge of realizing ethical-aware processes to dynamically cope with runtime conditions remains unaddressed. This paper proposes a methodology that leverages Large Language Models (LLMs) as planning components to dynamically shape human-robot collaboration. Our approach enables business processes to become comprehensively ethical-aware. We demonstrate the proposed methodology in an assistive care scenario through a proof-of-concept targeting a real robotic platform. The results show that, when provided with structured patient profiles and prompts, LLMs can generate personalized and context-sensitive action sequences that support ethical and adaptive human-robot collaboration.