Sheila Jasanoff

In the [YOU] on AI Field Guide

The cycle that began with [YOU] on AI confronts the reader with the Trivandrum training: twenty engineers whose job descriptions changed in a week, whose professional identities were reorganized around a tool they had not chosen and whose deployment they had not been consulted about. Jasanoff's framework names what Segal's honest narrative shows but does not fully theorize: the technical transformation and the social transformation were not sequential but simultaneous—co-produced. When Claude Code entered the engineering team's workflow, it did not merely change what the engineers could produce. It changed what an engineer was. The dissolution of specialist roles, the redistribution of authority between human judgment and machine output, the redefinition of what counted as expertise—these were not downstream effects of a primarily technical change. They were the change itself.

The governance gap that Segal identifies as one of the AI transition's defining structural features receives, from Jasanoff, a reframing that changes what kind of problem it is. The conventional telling treats the gap as a speed problem: technology moves fast, institutions move slowly. Jasanoff argues that this framing is itself a mechanism for delegating power—treating the governance gap as a natural fact rather than a design choice makes deference to technologists look rational. Her alternative: the gap is not a speed differential but a scalar mismatch. The decisions that must be made are global, instantaneous, and operating simultaneously in every domain of human activity. The institutions available to make them are national, slow, and organized around a sequential model—first the technology, then the assessment, then the regulation—that the AI moment has broken.

Jasanoff's sharpest intervention in the cycle's discourse concerns epistemic exclusion: the systematic silencing of certain kinds of knowledge through institutional design rather than censorship. The governance frameworks being built to address AI—the EU AI Act, the American executive orders, the corporate responsible AI frameworks—are almost entirely expert-driven. They classify risks using technical taxonomies. They impose requirements using instruments designed for a previous generation of technologies. They address what the technology can do and largely ignore what the technology does to the people who use it—what it costs in cognitive autonomy, professional identity, creative depth, relational presence. This knowledge is not absent. It circulates in Substack posts, in dinner-table anxieties, in the specific grief of the master calligrapher watching the printing press arrive. It is absent from governance, because governance institutions were built to process one kind of input, and this is not that kind.

The twelve-year-old who asks “What am I for?” possesses knowledge that no AI safety benchmark can capture. Jasanoff's project—and the cycle's most important governance question—is the construction of institutions that can hear her.

Origin

Jasanoff was born in Patna, India, in 1944, and trained first in law at Harvard before gravitating toward the nascent field of science and technology studies in the 1980s. Her doctoral and early research work centered on how regulatory agencies produce scientific knowledge about risk—how the FDA decides a drug is safe, how the EPA determines a chemical is dangerous, how the Nuclear Regulatory Commission establishes that a reactor design is acceptable. What she found was not what the standard account of regulation predicted: not scientists producing knowledge that regulators then apply to policy, but knowledge and policy being produced together, each shaping the other, in a process her later work would call co-production.

Her comparative turn, which produced the framework of civic epistemology, came from the recognition that different nations reached dramatically different conclusions about the same scientific questions—not because they had different science but because they had different ways of determining what the science meant and how it should translate into governance. Her comparative studies of biotechnology governance in the United States, Britain, Germany, and the European Commission—culminating in Designs on Nature (2005)—established that these differences were not incidental but systematic: expressions of deep cultural commitments about expertise, democratic legitimacy, and the appropriate distribution of risk.

The concept of technologies of humility arrived in a 2003 essay that has become one of the most cited works in science and technology studies. Jasanoff introduced it as the institutional counterpart to technologies of hubris—the quantitative risk assessments, cost-benefit analyses, and expert-dominated decision processes that characterize most technology governance. Technologies of hubris are not useless; they produce genuine knowledge. But they produce it within a framework that systematically overestimates what can be known and underestimates what cannot, transforming unmeasurable uncertainty into measurable risk and treating the transformation as an achievement rather than an assumption.

Her 2015 volume The Ethics of Invention and her contributions to the emerging field of AI governance brought the framework into direct confrontation with the technology industry's epistemic culture—its conviction that demonstration is evidence, that working prototypes are arguments, and that the people who build powerful things are the natural governors of those things. Jasanoff's counter-argument: competence within a domain does not confer authority over its consequences, and the consequences of AI radiate outward into workplaces and classrooms and dinner tables where no benchmark can follow them.

Key Ideas

Co-production. Co-production is the foundational thesis: knowledge and social order are made together, each constituting the other in real time. When a new capability enters a workflow, it does not merely add to what is possible; it reorganizes what the workflow is, who the people in it are, what counts as expertise, and how authority is distributed. The Trivandrum training was not a technical event with social consequences. It was a single co-produced event in which the technical and the social were inseparable from the first moment. Governance frameworks that separate them—addressing the technology in one policy and its social effects in another—are not governing the phenomenon. They are governing half of it.

Civic epistemology. Civic epistemology is the culturally embedded way a society produces and validates public knowledge. The Silicon Valley civic epistemology privileges demonstration: if you can build it and show it works, that constitutes sufficient justification for deploying it. The European civic epistemology is consensual and precautionary: knowledge about consequences must be established before deployment, not after. The Chinese civic epistemology treats AI governance as an expression of state capacity. None of these is simply correct; each captures something real and suppresses something equally real. AI governance that does not recognize these differences will consistently talk past itself in international forums.

Technologies of humility. Technologies of humility are institutional practices for governing under genuine uncertainty. Jasanoff identifies four components: framing (who defines the problem, and what does that definition include and exclude?), vulnerability (who is most exposed to harm, and how do they differ from those the technology was designed for?), distribution (who benefits and who bears the cost?), and learning (how do institutions detect their own errors and revise?). None of these questions can be answered by a safety benchmark. All of them are answerable by institutions designed to ask them.

The law-lag narrative. The law-lag narrative—technology moves fast, governance moves slowly, nothing can be done until after the damage is done—is, in Jasanoff's analysis, a mechanism for delegating power rather than a description of reality. Law and social institutions do not merely react to technology; they co-produce it. The patent system shaped the direction of industrial innovation for centuries. Environmental regulation reshaped industrial chemistry. The AI moment appears to confirm the law-lag narrative only if the analysis begins in 2022—treating everything that followed as governance playing catch-up to a fait accompli. The story begins much earlier, inside regulatory frameworks that made specific choices about data, intellectual property, and market structure whose consequences are now manifesting at scale.

Sociotechnical imaginaries. Every powerful technology arrives wrapped in a collectively held vision of the future it will create. Sociotechnical imaginaries are not predictions—they are blueprints that organize action, determine what counts as progress, and shape which consequences seem acceptable and which seem intolerable. The productivity imaginary sees AI as a capability amplifier democratizing creation. The existential risk imaginary sees AI as a potential extinction threat requiring extraordinary governance. The democratic imaginary—Jasanoff's own—treats AI as a constitutional question: what kind of society do we want, and how should AI be governed to serve those values?