Metrics for Good?: Quantification and Organizational Change in Sustainability and Responsible AI

• In Event: 40463 - Climate Change, Risk, and Markets

Abstract

Can metrics help organizations and their employees mobilize for positive social change? In this article, we compare two domains where for-profit corporations faced major public backlash: sustainability and Responsible AI (Artificial Intelligence). In both cases, quantification emerged as a key strategy to incentivize internal and field-level efforts. Drawing on two qualitative studies of sustainability and Responsible AI practitioners, we find that external pressures towards quantification were more institutionalized in the case of sustainability, which came with a tighter integration of metrics into internal operations compared to Responsible AI. At the same time, this tighter coupling of external pressures and internal practices in sustainability came with paradoxical consequences, namely that metrics became their own goal, decoupled from actual interventions when it came to climate change. In contrast, in Responsible AI, metrics remained largely decoupled from internal practices, leading workers to articulate broader public critiques of AI harms and industry norms through whistleblowing. We discuss the ramifications of these findings regarding the role of metrics and loose couplings in social and organizational change.

Authors

• Janna Zou Huang, University of California-Berkeley

• Angèle Christin, Stanford University

• Sanna J Ali, Stanford University