CONCEPT

Technologies of Hubris

Sheila Jasanoff's term for the governance instruments—quantitative risk assessments, cost-benefit analyses, expert-dominated decision processes—that produce genuine knowledge while systematically overestimating what can be known and underestimating what cannot.

Technologies of hubris are not useless. This is the crucial, carefully placed acknowledgment in Sheila Jasanoff's foundational 2003 essay, which introduced both the concept and its prescribed antidote, technologies of humility. Quantitative risk assessment is a genuine achievement. It forced disciplines of rigor onto questions about environmental contamination and pharmaceutical safety that had previously been answered by intuition or interest. Cost-benefit analysis made consequences legible in terms that legislatures and courts could act on. Expert-dominated decision processes concentrated relevant knowledge and protected it from political manipulation. These are real goods. The problem Jasanoff identifies is structural rather than motivational: technologies of hubris produce knowledge within a framework that transforms unmeasurable uncertainty into measurable risk, and treats the transformation as an achievement rather than an assumption. The untransformed residue—what cannot be assigned probabilities, what cannot be captured in a risk matrix, what is genuinely uncertain rather than merely unquantified—is treated as though it does not exist. For AI governance, this residue is not peripheral. The slow erosion of professional identity as expertise is commoditized, the gradual atrophy of cognitive capacities that are exercised only through friction, the displacement of human relationships by machine interactions that are more convenient and less real—these cannot be assigned probabilities. They are consequential. The governance frameworks built on technologies of hubris alone are not governing the AI transition. They are governing the part of it that fits in a spreadsheet.

Origin

Jasanoff coined “technologies of hubris” as the foil to “technologies of humility” in a 2003 essay in Minerva titled “Technologies of Humility: Citizen Participation in Governing Science.” The essay emerged from her decades of comparative research on how regulatory institutions in different national contexts produced knowledge about technology risks. In each case, she observed the same structural pattern: institutions built evidentiary frameworks that admitted certain kinds of knowledge and excluded others, and the excluded knowledge was systematically the knowledge of people who experienced the consequences of technology rather than the people who built it.

The concept of hubris here is not an accusation of arrogance directed at individual practitioners. Risk assessors and cost-benefit analysts are often genuinely humble people. The hubris is institutional and structural: it is built into frameworks that produce the appearance of certainty where genuine uncertainty exists, that transform the qualitative into the quantitative not because the transformation is valid but because the transformed product is actionable within existing institutional architecture. A risk matrix with probability estimates attached can be incorporated into a regulatory decision. A narrative account of how a technology is changing the felt quality of professional life cannot.

The AI moment has produced an extraordinary proliferation of technologies of hubris under the banner of “AI safety.” Safety benchmarks quantify a model's tendency to produce harmful outputs across specified categories. Alignment protocols measure the distance between a model's behavior and its designers' intentions. Red-teaming exercises systematically probe for failure modes. Each of these is a genuine technology of hubris: it produces real knowledge about specific, measurable properties of the system, and it systematically excludes the consequences that cannot be measured with the tools it employs.

Key Ideas

The probability-uncertainty substitution. The central operation of technologies of hubris is the substitution of measurable risk for unmeasurable uncertainty. Risk is a probability distribution over outcomes; uncertainty is the condition in which no such distribution can be reliably constructed. Real-world technology deployment almost always involves genuine uncertainty rather than quantifiable risk, because the consequences of a new technology arise from its interaction with a social order that cannot be fully modeled. Technologies of hubris transform this uncertainty into risk through a series of assumptions, and then govern the risk as though the assumptions were verified. The governance gap is partly a product of this substitution: institutions calibrated to manage quantifiable risk are not designed to handle genuine uncertainty.

The actionability constraint. Technologies of hubris persist not because they are epistemically superior but because they produce knowledge in a form that existing institutions can act on. A finding that “AI increases cognitive load by 23%” can be translated into a workplace standard. A finding that “sustained AI use is reorganizing the felt quality of professional identity in ways that accumulate below the threshold of conscious awareness” cannot. The evidentiary architecture of governance institutions selects for the actionable, and the actionable is the quantitative, and the quantitative is the measurable, and the measurable is almost never the most important.

The exclusion of experiential knowledge. Technologies of hubris are epistemically exclusive by design: they establish standards of evidence that the expert can meet and the citizen cannot. The engineer knows the specifications. The resident near the facility knows what it smells like at 3 a.m. and what it has done to her family's health. In every governance domain Jasanoff has studied, including AI, the former knowledge is admissible and the latter is not—not because the latter is less true but because the existing institutional architecture was built to process the former. The epistemic exclusion is structural, not intentional.

Contrast with technologies of humility. The prescriptive counterpart, technologies of humility, are institutional practices designed to operate inside genuine uncertainty rather than against it. They attend to framing (who defines the problem), vulnerability (who is most exposed to harm), distribution (who benefits and who pays), and learning (how institutions detect their own errors). The point is not to abandon technologies of hubris but to surround them with technologies of humility that supply what quantitative frameworks cannot: the knowledge of people who are experiencing the consequences, the recognition of what cannot be measured, and the institutional capacity to revise when reality diverges from the model.

Debates & Critiques

The productive debate about technologies of hubris is not whether quantitative risk assessment is legitimate—it is—but whether it should be the primary or even the dominant evidentiary framework for AI governance, where the most consequential consequences are qualitative and genuinely uncertain rather than quantifiable. Critics of Jasanoff's framework from within the risk analysis community argue that the alternative to quantitative frameworks is not wisdom and humility but the unchecked authority of whoever gets to define “experiential knowledge”—that the apparent openness of technologies of humility masks a new form of expert authority, the authority of the STS scholar who determines whose experience counts. Jasanoff's defenders note that this critique applies equally to quantitative frameworks: the apparent objectivity of the probability estimate masks the authority of whoever chose the probability model, defined the outcome categories, and established the threshold for acceptable risk. A second debate, more specific to the AI context, concerns the relationship between technologies of hubris and co-production: if technical knowledge and social order are produced simultaneously, then technologies of hubris are not merely epistemically limited but constitutive of a social order that privileges expert knowledge and marginalizes the knowledge of affected communities. The governance framework is not just inadequate; it is part of what it is supposed to govern.

Origin

Key Ideas

Debates & Critiques

Related Entries

Further Reading