Principles2Plan: LLM-Guided System for Operationalising Ethical Principles into Plans
Tammy Zhong, Yang Song, Maurice Pagnucco
TL;DR
The paper tackles the difficulty of operationalising abstract ethical principles in automated planning by introducing Principles2Plan, a human–LLM collaborative prototype that derives context-specific ethical rules from high-level principles and translates them into PDDL-Ethical for a classical planner. The authors describe a four-step interactive workflow (input, rule generation/review, prioritization, and plan generation) supported by an Ethical Rules Editor and a Code Editor, with transpilation to PDDL and execution in Fast Downward. They validate the approach using example domains (autonomous vehicles, elderly care, firefighting) and report preliminary metrics from a relevant evaluation model: Sentence-BERT similarity of 0.82 for generated rules and a code-generation success rate of 82.2%. The work demonstrates the feasibility of practical, transparent, ethics-informed planning and sets a foundation for future enhancements in human–LLM collaboration and iterative dialogue. Overall, Principles2Plan offers a concrete pathway to operationalise ethical principles in planning and invites further refinement to improve usability and performance.
Abstract
Ethical awareness is critical for robots operating in human environments, yet existing automated planning tools provide little support. Manually specifying ethical rules is labour-intensive and highly context-specific. We present Principles2Plan, an interactive research prototype demonstrating how a human and a Large Language Model (LLM) can collaborate to produce context-sensitive ethical rules and guide automated planning. A domain expert provides the planning domain, problem details, and relevant high-level principles such as beneficence and privacy. The system generates operationalisable ethical rules consistent with these principles, which the user can review, prioritise, and supply to a planner to produce ethically-informed plans. To our knowledge, no prior system supports users in generating principle-grounded rules for classical planning contexts. Principles2Plan showcases the potential of human-LLM collaboration for making ethical automated planning more practical and feasible.