Power and Public Values

In “What Do We Want From AI?” (Issues, Spring 2026), John P. Nelson takes stock of the many ways that artificial intelligence is affecting our lives and societies, and puts forwards the lens of “public values” to help orient our decisionmaking on such consequential issues. His prescriptions and proposals are sound. They invoke the importance of democratic deliberation, negotiation, and experimentation. They demonstrate concrete ways in which policymakers, private firms, and other stakeholders—anywhere from the court room to the board room—could leverage public values to “get a head start” and ultimately shape AI in the ways we collectively want. Indeed, it would not be surprising if large majorities of the public supported Nelson’s specific policy proposals surrounding deepfakes, data centers, and AI labor disruption.

Yet key AI decisionmakers in both the policy and private sectors are failing to act accordingly. This suggests the underlying problem is not the challenge of garnering agreement over complex ethical trade-offs, nor of articulating societal goals through the lens of public values, but something more mundane—power.

That is, we often know what matters, what the public cares about, what we should do. We just don’t do it. We lack demand for public values. One uncharitable interpretation is that adopting a public values lens would be “merely” repackaging sibling strategies such as AI ethics or responsible innovation that could likely sufficiently guide ethical development and governance. The bigger issue is thus not that we are unaware of public values, but rather that private values remain more functionally important in society.

One strategy often proposed to counter this gap is increased public participation in AI. As Nelson recognizes, uncovering public values on its own, such as through public opinion polling, is insufficient. In response, he proposes several ideas that are prominent in civic and academic circles, such as public forums, scenario planning workshops, interdisciplinary thinking in AI design and governance, and deliberative approaches generally. These strategies, all worth pursuing, could help boost public participation.

This raises a chicken-and-egg problem. Citizens need political efficacy, desire, and capacity to leverage participatory structures toward change—but the structures needed to build these capacities and empower citizens in the first place do not yet exist at scale—in the United States at least. Consequently, it seems difficult to “pull ourselves up by our bootstraps” and foster robust participation that could disrupt the political terrain and promote meaningful integration of public values. If anything, the quality of public participation and civic discourse appears to be flailing: In an interconnected world where people have abundant access to information and shiny new tech tools to promote democratic engagement, individuals are also becoming more disenfranchised, disaffected, addicted, and polarized. While laudable then, I therefore fear the “participatory solution to AI governance” is proving insufficient.

Fundamentally, private values and closely held power are still the kingmakers. Personally, I’m sold on the importance of public values, participatory AI approaches, and even the specific policy proposals Nelson offers. Yet until our institutions and leaders come to demand public values—or until we insist upon them—we may not get what we want from AI.