PolicyPad: Collaborative Prototyping of LLM Policies

An interactive system supporting domain experts in creating policies governing LLM behavior, drawing inspiration from UX prototyping methods like heuristic evaluation and storyboarding. Enables policy designers to work together in real time while simultaneously testing how proposed policies affect LLM outputs through concrete usage scenarios. Evaluation with 22 domain experts in mental health and law showed strengthened collaborative teamwork, faster feedback cycles, and innovative policy ideas, representing participatory paths for advancing AI alignment and safety.