The AI Power Disparity Index: Toward a Compound Measure of AI Actors’ Power to Shape the AI Ecosystem

Abstract

The rapid advancement and adoption of Artificial Intelligence (AI) is reshaping power dynamics among key AI stakeholder groups. Capturing this shift requires a framework to assess various dimensions of power held by different players in the AI ecosystem. While previous work has explored AI power dynamics from theoretical and philosophical perspectives, little attention has been given to concretely measuring who is involved in and impacted by the shifting power dynamics, which dimensions of power are evolving over time, and to what extent. In this paper, we conceptualize the AI Power Disparity Index (AI-PDI): a composite indicator designed to measure and signal the changing distribution of power in the AI ecosystem. Drawing on interdisciplinary theories, we propose to aggregate several key dimensions of power in the index, including its bases, means, scope, and degree. While the primary contribution of this work is to conceptualize the index, we take an initial step toward operationalizing it: inspired by the Delphi method, we recommend a collaborative, expert-driven methodology to determine the specific form and content of the index. We envision the AI-PDI to serve as a critical resource for policymakers, researchers, and civil society to track evolving power dynamics around AI, encouraging informed dialogue and guiding long-term strategies for democratic and transparent development and use of the technology.