Merritt Roe Smith

In the [YOU] on AI Field Guide

[YOU] on AI opens with the river metaphor: intelligence has been flowing for 13.8 billion years, and the current moment is a threshold in that flow. The metaphor carries unmistakable determinist undertones—a river is a natural force, it does not negotiate, it flows according to the laws of physics. Smith's framework is the corrective. The river flows, yes—but the beaver redirects it. The dam creates a pool, the pool creates an ecosystem, and the ecosystem supports forms of life the unimpeded river would have swept away. The institutional response to a technology is the dam. And the choices being made right now, during what Smith's framework identifies as the formative period of the AI transition, will constrain possibilities for decades.

Smith stands in the cycle's gallery as the historian who refuses the temptation—in either its optimistic or apocalyptic mode—to treat the AI transition as a singular, inevitable force following its own trajectory. His comparative method reveals the mechanism by which the same technology produces divergent outcomes in different institutional contexts. The Harpers Ferry Armory and the Springfield Armory received the same machines, the same federal mandate, the same blueprints—and diverged dramatically, because their organizational cultures differed. The AI deployment happening right now in Trivandrum and in Tokyo and in Paris and in Lagos is the same technology landing on different institutional soil, and the outcomes will differ accordingly.

The cycle's concept of the soft determinist position—the beaver that does not stop the river but redirects it—is precisely the intellectual stance Smith's lifetime of research vindicates. It is more honest than the triumphalist account that treats the transition as an unambiguous gain, and more empowering than the elegiac account that treats it as an unstoppable displacement. It says: the technology is genuinely powerful, the constraints it imposes are real, and within those constraints the specific future remains open to institutional choice. The choice is being made now. The window is closing.

Where Meredith Whittaker focuses on who owns the AI infrastructure and Michael Crawford focuses on what the practitioner loses when friction is designed away, Smith provides the historical depth: every major technological transition in modern history has been shaped, in its specific form and its specific distribution of costs and benefits, by the institutional arrangements that surrounded it. The transitions that produced broadly shared prosperity were not the ones where the technology was gentler. They were the ones where the institutions were stronger.

Origin

Smith built his reputation at MIT through an unlikely subject: the federal armories at Springfield, Massachusetts, and Harpers Ferry, Virginia, where the U.S. Army developed interchangeable-parts manufacturing in the early nineteenth century. The research was archival, painstaking, and unfashionably empirical. His 1977 book Harpers Ferry Armory and the New Technology documented the decade-long resistance of skilled craftsmen to the introduction of precision manufacturing—a resistance that was not irrational but institutionally rational, grounded in a genuine understanding of what the reorganization would cost them. The book established a pattern Smith would document repeatedly: that workers who resist technological change are often the people who understand its costs most clearly, and that their resistance is an essential resource for the institutional design that determines whether a transition produces broadly shared benefit or concentrated gain.

His edited volume Military Enterprise and Technological Change: Perspectives on the American Experience (1985) extended this analysis across multiple domains, documenting the military as the primary engine of American technological development and the carrier of specific institutional values—efficiency, standardization, scalability, control—into civilian applications that those values were not always suited to serve. The military-industrial origins of computing, which Smith's contributors traced, provide the institutional genealogy that explains why AI's architecture, optimization criteria, and evaluation metrics look the way they do.

Smith's later work addressed the determinist question directly. His essay in the volume he co-edited with Leo Marx, Does Technology Drive History? The Dilemma of Technological Determinism (1994), gave conceptual shape to the soft determinist position he had always implicitly occupied: technology constrains the range of possible futures without determining which specific future materializes. The answer to his title question was—it depends on the institutions.

Key Ideas

Soft determinism. Smith's foundational intellectual contribution is the distinction between hard and soft determinism. Hard determinism says technology determines specific outcomes regardless of institutional context—producing either triumphalism or fatalism, both forms of passivity. Soft determinism says technology constrains the range of possible outcomes while leaving the specific outcome genuinely open to institutional choice. The Springfield-versus-Harpers-Ferry comparison is the paradigmatic demonstration: identical technologies, different institutional cultures, dramatically divergent outcomes. Not the technology. The institutions.

Path dependence and the formative period. Smith's most practically urgent claim is temporal: early choices carry disproportionate weight. The decisions made during the formative period of a technological transition—about architecture, business models, regulatory frameworks, educational responses—accumulate the institutional investments that make later deviation increasingly costly. Path dependence is not fate; it is the accumulation of choices into structures that constrain future choices. The AI transition is in its formative period now. The window for institutional innovation is narrowing with each month of unmediated adoption.

The myth of the neutral tool. Technologies are not neutral instruments whose effects depend entirely on how users choose to employ them. They embody the values of the institutions that produced them, values that persist long after the original institutional context is forgotten. The precision manufacturing tools at Springfield were designed to replace craft knowledge with standardized procedure—and that design was the most consequential decision, made before any worker touched the machine. The large language models now reshaping knowledge work embody the values of the research environments that produced them: helpfulness, fluency, breadth, user engagement. These values are not neutral in their effects on the humans who use them.

The Luddites were right about the costs. Smith's insistence on documenting worker resistance with the same archival rigor he applied to documenting innovation is not merely a historiographical preference. It is a substantive claim: the people displaced by new technologies are not irrational obstacles to progress but thinking, strategically capable actors whose responses shape the transition's course. The Luddite response was rational, grounded in genuine understanding of the economic dynamics at play. The craftsmen at Harpers Ferry who resisted mechanization were not wrong about what the reorganization would cost them. They were unprotected.

Institutional mediation is the mechanism. The outcomes of technological transitions are determined not by the technology itself but by the institutional arrangements that mediate between the technology's capabilities and its social effects. The Factory Acts established the principle, the labor protections built on it, and the social insurance programs extended it: that the deployment of powerful technology in domains affecting human welfare is a matter of legitimate public concern requiring institutional attention. Each generation of institutional response was imperfect, late, and the product of political struggle. Each was also decisive in determining whether a technological transition produced broadly shared benefit or concentrated gain at dispersed expense.