Five Ps: Leverage Zones Towards Responsible AI

Most current Responsible AI efforts are "low-leverage" fixes: The paper finds that the majority of interventions focus on short-term solutions like tweaking algorithms and updating parameters, rather than addressing the foundational structures and purposes that govern AI systems—meaning current efforts may not create lasting change.

Responsible AI risks becoming corporate marketing: Without more effective interventions targeting root causes, the concept of Responsible AI could become merely a buzzword used by companies for marketing purposes rather than a genuine commitment to systemic change.

A new framework is needed to evaluate intervention effectiveness: The paper proposes the "Five Ps" framework, adapted from Systems Thinking, to help identify which interventions actually contribute meaningful change versus those that only provide superficial improvements.

High-leverage interventions require challenging foundational assumptions: Real progress requires not just fixing technical problems but questioning the underlying purposes and structures upon which AI systems are built—a more difficult but necessary approach than current practices.

Transdisciplinary collaboration is essential: The framework is designed to scaffold question-asking across different disciplines, suggesting that solving Responsible AI challenges requires perspectives beyond computer science and engineering alone.