Archon Fung

In the [YOU] on AI Field Guide

[YOU] on AI advances the proposition that the most consequential error of the Luddites was not their fear of new technology but their withdrawal from the spaces in which its deployment was being governed. Stay in the room. Build the dams. The people who shape the future are the people who show up. Fung's response to this proposition is precise and unsentimental: the injunction presupposes that a room exists, that it is accessible to the people who need to be in it, and that presence in the room translates into influence over the decisions made there. Each of these presuppositions, examined against the evidence of how AI governance actually operates, proves false. The framework knitters of Nottinghamshire did not choose to withdraw from governance. They were excluded from governance by institutional arrangements that reserved participatory power for a narrow elite. The dams that eventually distributed industrialization's gains more broadly were built not by individuals who chose to engage but by social movements that fought, at enormous cost, to restructure the governing institutions themselves.

Fung reads the Trivandrum experiment at the center of [YOU] on AI—twenty engineers encountering AI tools in a structured, mentored, economically secure environment—as one of the clearest natural experiments in empowered participatory governance available in the current period. The process was accessible: engineers had dedicated time and co-location. It was deliberative: discoveries became collective through shared engagement. It was consequential: findings directly shaped organizational deployment decisions. But Fung identifies the experiment's structural limit immediately: it succeeded because one organizational leader chose participatory governance. A governance model that depends on the accident of individual leadership is not a governance model. It is a lottery. The challenge is to convert the Trivandrum model from an exceptional instance of enlightened practice into a standard institutional requirement.

The fishbowl condition—the structural state in which affected populations can observe AI governance processes but cannot influence them, separated from decision-makers by a barrier that is transparent but impenetrable—is Fung's precise adaptation of the cycle's fishbowl metaphor to the democracy problem. Unlike the fishbowls of disciplinary limitation that the cycle describes, the governance fishbowl is actively engineered: public comment periods that privilege technically literate respondents, stakeholder sessions convened by the very entities under scrutiny, legitimation mechanisms that produce the appearance of inclusion without its substance. Consultative theater is not harmless; it inoculates institutions against future demands for real participation by allowing them to claim that participation has already been tried.

The recursive trap Fung identifies—AI shaping the governance environment within which AI governance occurs—connects directly to the cycle's treatment of attentional ecology. The same AI capabilities that manipulate individual voters can subvert the regulatory comment periods designed to channel citizen voice: generating millions of synthetic comments advancing a given policy position, drowning genuine public input in synthetic noise. The fishbowl does not merely exclude affected populations; it is now equipped with technology that can simulate their inclusion. Fung's framework identifies deliberative participation by affected populations as the governance mechanism most structurally resistant to this form of degradation: face-to-face deliberation, random selection, structured information provision, and binding authority over outcomes are the features that protect democratic governance from the specific threats AI poses to it.

Origin

Archon Fung is the Winthrop Laflin McCormack Professor at Harvard Kennedy School, where he has taught since 1999. His foundational research on participatory governance was conducted across multiple continents and governance domains, synthesized in Deepening Democracy (2003, with Erik Olin Wright) and Empowered Participation (2004). The latter examined Chicago's Empowerment Zone and community policing programs as experiments in genuine empowered participation—processes that met all three of his conditions and produced measurable improvements in governance outcomes compared to expert-only alternatives.

The June 2023 encounter with AI governance that became central to his work on the AI transition came when Senator Josh Hawley asked OpenAI's CEO whether AI language models could be used to manipulate voters through personalized, adaptive, one-on-one persuasion at a scale no human campaign could achieve. Sam Altman said yes, he was concerned about exactly that. Within weeks, Fung and Lawrence Lessig published the Clogger analysis, constructing a thought experiment around an AI designed to maximize the probability that its candidate wins an election with no regard for truth. The analysis appeared across Scientific American, Salon, Asia Times, and dozens of newspapers. Its central insight was not that AI could produce bad electoral outcomes—a familiar warning—but that an AI arms race in political persuasion would produce outcomes determined not by the quality of ideas or the preferences of citizens but by the relative effectiveness of competing manipulation machines.

Fung has been a co-director of the Ash Center for Democratic Governance and Innovation at Harvard and a founding contributor to the GETTING-Plurality Research Network, which connects technology governance to democratic theory. His December 2024 workshop on AI and democracy movements confronted the recursive trap directly: democracy movements worldwide were experiencing a historic decline in their capacity to challenge autocratic governments, partly attributable to the AI advantage those governments possessed in surveillance, censorship, and propaganda. The workshop's finding was stark: the technology landscape was widening the gulf between democracy movements and their adversaries, and the relatively slow AI adoption by those movements was the gap that mattered.

Key Ideas

Empowered participatory governance. Fung's central framework identifies three conditions that must be simultaneously satisfied for participation to produce genuine democratic governance. Accessibility means barriers to entry are low enough that affected populations can participate without bearing disproportionate costs. Deliberation means participants engage with relevant information, hear competing perspectives, and refine their positions through structured dialogue. Consequence means participatory outcomes exercise genuine influence over actual decisions. Any two without the third degrade into a distinct pathology: informed people with no governance weight, binding decisions made on uninformed gut reactions, or thoughtful elite deliberation that systematically excludes the people most affected.

The fishbowl condition and consultative theater. The fishbowl condition is the structural state in which affected populations can observe governance processes but cannot influence them. Consultative theater is the result: the performance of inclusion that produces the appearance of democratic legitimacy without its substance. It is actively destructive because it consumes attention, betrays expectations, and erodes the trust that genuine participation requires—while allowing institutions to claim that participation has already been tried.

The Clogger thought experiment and the recursive trap. Clogger—an AI designed to maximize electoral victory through personalized persuasion with no regard for truth—reveals not just that AI can produce bad electoral outcomes but that AI threatens the conditions under which democratic governance of AI is possible at all. A technology being regulated is actively reshaping the conditions under which regulation occurs. This recursive trap cannot be broken by better algorithms or more transparent AI systems. It can only be broken by governance institutions that operate on principles resistant to algorithmic subversion: face-to-face deliberation, random selection, and binding authority over outcomes.

Minipublics and the silent majority. Minipublics—small, demographically representative bodies of randomly selected citizens who deliberate on specific governance questions under conditions designed to produce informed, considered judgments—are the governance mechanism Fung identifies as most resistant to the specific forms of democratic degradation AI produces. The silent middle that [YOU] on AI identifies—the population that experiences the AI transition with informed ambivalence but has no institutional channel through which to express it—is the constituency that random selection and deliberative structure are specifically designed to access. It is also a knowledge resource: the practical wisdom of workers, parents, and citizens that expert-only governance systematically fails to include.

The Transition Deliberation Committee. At the organizational level, Fung proposes the Transition Deliberation Committee—a standing body within organizations deploying AI systems, with formal authority over specified deployment decisions including the pace of implementation, training program design, and the allocation of productivity gains. The design adapts European works councils and German codetermination structures, with one critical Trivandrum-derived addition: structured hands-on engagement with the AI systems being governed, so that practical knowledge from direct experience informs deliberation rather than remaining inaccessible to it.