James C. Scott

In the [YOU] on AI Field Guide

The cycle that began with [YOU] on AI names the “silent middle”—the majority of practitioners, workers, and citizens who feel the AI transition’s weight and cannot articulate their experience in the binary terms the public discourse rewards. Scott spent his career giving the silent their vocabulary. The concept of métis—the practical, local, embodied, contextually adapted knowledge that practitioners build through sustained engagement with their domain—is precisely the knowledge the silent middle possesses and that the governance frameworks being drafted in boardrooms and ministry offices do not solicit. The consequence, if Scott’s historical pattern holds, is the AI equivalent of Waldsterben: a managed landscape that looks rational from above and is dying from below.

His four conditions for catastrophe—high modernist ideology, institutional power sufficient to impose the plan, a prostrate civil society unable to resist, and the absence of feedback mechanisms that would reveal failures in time to correct—are not all fully assembled in the AI transition yet. But they are being assembled. The comprehensive AI strategies, the corporate governance frameworks, the academic integrity policies are the ideology. The platform power of a handful of companies is the institutional power. The silent middle is the civil society in formation. The absence of channels through which practitioner métis can reach the people making decisions is the failing feedback mechanism. The window for intervention is open. Scott’s record is a warning about how quickly it closes.

Where James Madison asks how power should be arranged when it cannot be trusted, Scott asks what kind of knowledge is required to exercise power without destroying what it governs. The two questions are complements. Madison’s structural answer and Scott’s epistemic answer together specify what adequate AI governance would require: structures with teeth, and a genuine solicitation of the knowledge that only practitioners possess.

The concept of weapons of the weak—the everyday resistance strategies of people who cannot fight openly—is directly visible in current AI adoption dynamics: the foot-dragging, the false compliance, the feigned ignorance, the professional culture that quietly delegitimizes the early adopter. Scott’s analysis is honest about their limits: they slow the transition and preserve the resister’s dignity, but they do not change structural outcomes. The most valuable thing a resister possesses is not her refusal but her métis—her knowledge of where the tools fail, what the dashboard misses, what the plan cannot see. That knowledge is useful only if it remains inside the system, available to influence the people making decisions.

Origin

James C. Scott was born in 1936 and spent most of his career at Yale, where he held the Sterling Professorship and founded the Agrarian Studies Program. His intellectual formation was shaped by two sustained engagements: with Southeast Asian peasant politics, which he studied through fieldwork in Malaysia and which produced The Moral Economy of the Peasant (1976) and Weapons of the Weak (1985); and with the structural analysis of how states see and simplify the complex social realities they attempt to govern, which culminated in Seeing Like a State (1998). The Malaysian fieldwork was crucial: living in the village of Sedaka in the early 1980s and watching the Green Revolution play out from below gave Scott an irreplaceable experience of the gap between the planner’s knowledge and the practitioner’s. The gap was not a matter of the planners being wrong about the yields; they were often right. It was a matter of the planners not knowing what they did not know—specifically, the intricate social and ecological knowledge that the peasants had built through generations of engagement with a specific, resistant environment.

Later works extended the framework: Domination and the Arts of Resistance (1990), The Art of Not Being Governed (2009), and Against the Grain: A Deep History of the Earliest States (2017), which traced the legibility project back to the earliest attempts at administrative control of complex populations. Throughout his career Scott remained interested in the practical knowledge of people at the periphery of power—the farmer, the peasant, the craftsperson—and suspicious of the confident rationality of people who govern from above. He was not anti-technology or anti-modernity; he was anti-hubris, and the distinction is exactly the one the AI transition requires.

Key Ideas

Métis: the knowledge that resists formalization. Scott adopted the Greek concept for the practical, local, embodied, contextually adapted knowledge that cannot be extracted from the context in which it was built without being destroyed. The sailor who reads the sea by its color and the pattern of its swells; the midwife who feels the position of the child through the mother’s abdomen; the engineer who can feel when a codebase is fragile—all possess métis. It is real knowledge, as real as any theorem, but it is too local to aggregate, too embodied to document, and too dialogical to capture in a policy framework. It is also exactly the knowledge that determines whether an AI deployment works or fails in the specific context where it is deployed.

High modernism and its blindness. High modernism is not ignorance; it is a particular kind of intelligence—the intelligence of the administrator, the systems architect, the planner—applied with such confidence that it overrides the métis of the people who actually inhabit the systems being redesigned. The Prussian foresters were excellent scientists. Their knowledge of trees was genuine. Their blindness to the mycorrhizal network, the nutrient cycling, the ten thousand interactions that made the forest a living system rather than a timber farm, was not stupidity but the structural consequence of their position: they could see what their models could measure and could not see what those models left out. High modernism pervades AI governance: the comprehensive strategies, the risk taxonomies, the ethics frameworks, all built by people whose technical knowledge is genuine and whose knowledge of what practitioners actually experience is absent.

Legibility as the precondition and the trap. A state must make its territory legible to govern it; it cannot tax what it cannot count. Legibility is not bad. The trap is the substitution of the legible representation for the complex reality it was designed to capture. The productivity dashboard is a legibility device; it captures what is measurable and omits what is not; and when organizations optimize for the dashboard rather than the underlying reality, the unmeasured things—architectural judgment, the métis of knowing where the AI fails—are degraded and eventually lost.

Weapons of the weak and their limits. People who cannot fight openly fight indirectly: foot-dragging, false compliance, feigned ignorance, pilfering, character assassination. These tactics are rational, not futile, and Scott honored them for what they are. But they are substitutes for power, not expressions of it. They slow the transition; they do not change its direction. The same practitioner who uses weapons of the weak possesses the métis that effective governance most needs; the task is to convert that resistance into construction, to find institutional channels through which practitioner knowledge can enter the governance conversation rather than remain as the quiet rejection it currently is.