‘Governance by Design’ for AI

Close up photo of Kenneth A. Bamberger smiling and wearing glasses and a light red collared shirt.

Recentering Public Values In AI Governance: Examples From The Biden Administration,” a new Berkeley Technology Law Journal article by Berkeley Law Professor Kenneth A. Bamberger and UC Berkeley School of Information Professor Deirdre Mulligan analyzes the Biden-Harris administration’s AI policies through a “governance by design” framework they developed. 

That approach puts public values, sectoral expertise, and participatory policymaking at the center of decisions to regulate through technology, prioritizing human and public rights while developing risk-assessment tools; expanding the expertise of federal agencies by hiring from and collaborating with the relevant sectors; and keeping policy debaets public-facing by emphasizing engagement and potential impact throughout the development of AI systems. 

Bamberger and Mulligan — both faculty co-directors of the Berkeley Center for Law & Technology — cite three examples of administration policies that reflected the governance by design principles: 

  • Its layered regulatory strategy that empowers sectoral agencies to safeguard human rights and public safety in AI use
  • The expansion of AI and rights-based expertise within government and the establishment of collaborative structures for risk management and evaluation
  • The institutionalization of practices that surface and interrogate the normative assumptions embedded in AI systems, while scaffolding public participation throughout their lifecycle

“We argue that together these initiatives offer an alternative to prevailing AI governance debates — particularly the dichotomy between risk-based and rights-based approaches, and the call for a centralized AI regulator,” Bamberger and Milligan write. “Instead, such governance-by-design provides a field-centric model that leverages existing institutional capacities, protects democratic norms, and re-centers the public in the often-private domain of AI development. It offers a durable, epistemically responsible framework for regulating AI systems in a way that supports both human rights and legitimate democratic governance.”