Leverage zones in Responsible AI: towards a systems thinking conceptualization

This paper argues that current Responsible AI interventions haven't sufficiently addressed the root causes of AI problems. It proposes a novel approach using 'leverage zones' from systems thinking literature to plan and experiment with interventions. A conceptual framework called the Five Ps (Problem, Parameter, Process, Pathway, Purpose) is introduced to categorize interventions, from low-order fixes (e.g., algorithmic tweaks) to high-order changes (e.g., redefining system purpose). The framework is presented as a scaffold for transdisciplinary discussion to improve Responsible AI outcomes.

Ehsan Nabavi, Chris Browne