Sustainability-by-Design: Public Values and Artificial Intelligence

• In Event: Technology Diffusion, Industry, and Global Competition

Abstract

As governments rapidly adopt artificial intelligence (AI) across administrative functions, debates in the public administration literature about its use have been largely shaped by four public values: accountability, equity, efficiency, and robustness. Despite growing evidence of AI’s environmental footprint, environmental sustainability remains mostly absent from public values discussions of AI in public administration and policy scholarship. Although sustainability is recognized as a public value in the public administration literature, it has not been meaningfully incorporated into assessments of public sector AI use, leaving a key public value largely ignored. In comparison, environmental sustainability makes up a core value in discussions of AI in the computer science literature.

This theoretical and conceptual paper examines whether sustainability must be treated as a core public value in public sector AI adoption and governance. It does so by considering how governments can govern AI systems more effectively through the inclusion of environmental concerns in governance processes. We conduct two parallel scoping literature reviews to outline the importance of considering sustainability in public value assessments. The first review assesses the literature on AI in government through a public values lens. The second analyzes AI research in computer science focusing on “greening” AI applications. This cross-disciplinary approach allowed us to identify conceptual gaps in public administration scholarship. We draw from these two reviews to highlight the importance of incorporating sustainability into discussions of AI in government and public values.

Our review reveals four core public values guiding discussions around its adoption in the public sector: accountability, equity, efficiency, and robustness. While sustainable AI has received limited attention in the public administration literature, the technical literature in computer science shows how sustainable AI practices can support these four public values. Our analysis observes that sustainable AI practices, such as model optimization and lifecycle planning, can reinforce these public values simultaneously. Sustainability can improve accountability by increasing transparency around computational costs. It supports equity by lowering barriers to technological access. It also strengthens robustness by reducing the risks of long-term environmental instability. Further, our findings from the sustainable AI literature suggest that optimizing AI systems for environmental efficiency does not necessarily reduce accuracy or performance, two key aspects of bureaucratic efficiency. ​We conclude that sustainability must be recognized as an important public value guiding public sector activity related to AI adoption and deployment. The integration of sustainability into public sector AI governance discussions can help governments balance competing value commitments while addressing emerging environmental concerns. Our paper concludes by proposing a research agenda for embedding sustainability into future AI and other large technological computing technologies as they develop in the public sector.

Author

• Nicolas Ferraiuolo, Arizona State University
  Presenting Author