Epistemic Exclusion

In the AI Story

Jasanoff documented epistemic exclusion across regulatory domains. In environmental governance, the experiential knowledge of communities living near pollution sources — 'the water tastes different, the children are sick more often' — was inadmissible in regulatory proceedings requiring peer-reviewed studies with statistical significance. By the time quantitative evidence caught up with experiential knowledge, the damage had often been done. In pharmaceutical regulation, patient knowledge about side effects and quality-of-life impacts was subordinated to clinical trial data, even when the trials measured outcomes patients did not value and ignored outcomes they did. The pattern was structural: institutions designed to process expert knowledge systematically excluded other forms, not because they were hostile to those forms but because they lacked the mechanisms to incorporate them.

The AI governance landscape reproduces this pattern with unusual clarity. The silent middle that Segal identifies — people who feel both exhilaration and loss but cannot find a place in a discourse rewarding clean narratives — is an epistemically excluded population. They possess the most accurate knowledge of the AI transition: what it feels like, day by day, to operate at the boundary between human judgment and machine capability. But their knowledge is qualitative, contradictory, and expressed in narrative rather than metrics. It does not fit the evidentiary architecture of governance institutions, and so it is treated as anecdote — acknowledged in opening remarks, absent from the analysis that follows.

The exclusion compounds through institutional design. When qualitative evidence is systematically excluded, governance frameworks develop around quantitative evidence alone. The frameworks become optimized for what they can measure (adoption rates, productivity gains, benchmark scores) and blind to what they cannot (identity erosion, meaning displacement, the slow transformation of what it feels like to know something). The institutions that produce governance knowledge — policy research centers, regulatory agencies, commissioned studies — learn to produce the evidence the governance architecture can process. They study what is measurable and ignore what is not, not from malice but from institutional incentive.

Jasanoff's prescriptive response is institutional redesign. Governance institutions must be built to process multiple forms of evidence — quantitative and qualitative, expert and experiential, predicted and emergent. This requires not merely collecting qualitative data but treating it as evidence: according it weight in decisions, designing analytic methods that can integrate it with quantitative findings, and training governance professionals to evaluate narrative accounts with the same rigor they apply to statistical analyses. The redesign is difficult and requires confronting the institutional culture that treats numbers as objective and stories as subjective — a culture Jasanoff has spent her career demonstrating is itself a political choice masquerading as epistemological necessity.

Origin

The concept synthesizes Jasanoff's co-production framework with the feminist epistemology tradition (particularly Iris Marion Young's analysis of structural injustice and exclusion) and the science studies literature on lay expertise (especially Brian Wynne's work on Cumbrian sheep farmers whose local ecological knowledge was dismissed by radiation experts after Chernobyl). Jasanoff's distinctive contribution was to show that epistemic exclusion is not a failure of particular institutions but a structural feature of how modern governance institutions were designed — and that addressing it requires redesign, not better implementation.

Key Ideas

Exclusion is structural, not malicious. Governance institutions exclude certain knowledge forms through evidentiary standards and procedural rules, not through deliberate silencing — making the exclusion invisible to the excluders and unchallengeable by the excluded.

What is excluded is often what matters most. The experiential knowledge of identity transformation, meaning displacement, and relational costs — knowledge governance institutions cannot process — is frequently more consequential than the quantifiable risks institutions are designed to address.

Exclusion produces illegitimacy. When affected communities recognize that their knowledge has been excluded from governance, they experience the resulting decisions as imposed — technically competent perhaps, but democratically inadequate.

Inclusion requires institutional redesign. Incorporating diverse knowledge forms is not a matter of better public relations or more consultation; it requires rebuilding governance institutions to process evidence they were not designed to handle.