The cycle that [YOU] on AI inaugurates is itself an attempt to interrupt a self-fulfilling prophecy. The book’s central argument—that large language models are amplifiers of human judgment rather than substitutes for it—is not merely descriptive. It is, in Merton’s precise terms, a definition of the situation designed to evoke behavior that makes the definition true. When the engineer in Trivandrum believes her expertise will become more valuable, not less, when amplified by AI tools, she engages with the tools in ways that produce the hybrid competence that confirms the belief. The prophecy of enduring value is self-fulfilling in exactly the same structural sense as the prophecy of obsolescence—but the reality it constructs is one of expansion rather than atrophy.
The senior architect who described himself as “a master calligrapher watching the printing press arrive” was not primarily making a statement about skills. He was constructing a definition of the situation—one in which his accumulated expertise had become an artifact of a superseded paradigm. That definition was self-fulfilling: once he accepted it, his behavior became consistent with it (withdrawal, reduced investment in craft, resistance to engagement with the tools), and the behavior produced the diminished professional position the definition predicted. The alternative definition—that the master calligrapher’s embodied judgment is precisely what the printing press operator needs to direct—was equally available and equally self-fulfilling. The outcome depended not on the technology but on which definition structured behavior.
Merton’s analysis of manifest and latent functions illuminates a dynamic the productivity narrative systematically conceals. Organizations that cut mentorship programs because they believe AI will substitute for human expertise are serving a latent function—dependency reduction—alongside the manifest function of cost reduction. The latent function operates regardless of whether the manifest belief is accurate. And if the manifest belief proves false—if human judgment retains its value and the tools require expert direction to produce expert output—the latent consequence (eroded mentorship infrastructure) will have already destroyed the conditions under which expert judgment develops. The prophecy will have fulfilled itself not through the technology’s actual capabilities but through the institutional response to the belief about those capabilities.
The Matthew Effect—Merton’s 1968 finding that advantage compounds in social systems through feedback loops between position and returns—predicts the distribution of AI’s benefits with uncomfortable precision. The developer in Lagos and the engineer at Google both have access to Claude. The Google engineer’s institutional infrastructure—deployment pipelines, investor networks, brand credibility, proximity to capital—multiplies the value of her AI-assisted output in ways the Lagos developer’s context cannot. The floor rises for both. The ceiling rises faster for the one who was already near it. Democratization of access does not equal democratization of benefit, and the Matthew Effect explains why.
Born Meyer R. Schkolnick in 1910 in South Philadelphia—the son of Eastern European Jewish immigrants—Merton transformed himself through exactly the kind of developmental passage his work would later illuminate: the adoption of a professional name, the movement from working-class origins to Harvard graduate school, the construction of an identity through intellectual achievement rather than inherited position. He spent the bulk of his career at Columbia University, building the Department of Applied Social Research into the most empirically rigorous institution in American sociology, while simultaneously developing the theoretical frameworks that would define the discipline.
The “Self-Fulfilling Prophecy” essay of 1948 drew on W. I. Thomas’s earlier theorem but sharpened it into something analytically new: not merely that beliefs have consequences but that the consequences of false beliefs can validate those beliefs through circular causation. Merton demonstrated the mechanism across racial discrimination (white employers who believe Black workers are unreliable refuse to hire them; the refusal denies employment experience; the denial produces the work history that “confirms” the original belief), economic crises (the bank run), and educational outcomes. The essay became one of the most cited works in the history of social science, and its central concept entered common usage so completely that its origin is now frequently forgotten.
Merton’s sociology was defined by two complementary commitments that most of his contemporaries treated as incompatible: rigorous empirical specificity (his “theories of the middle range”—specific, testable propositions about bounded social phenomena) and structural analysis (the insistence that individual behavior, however locally rational, is shaped by structural conditions that the individual cannot fully see). Unanticipated consequences, latent functions, and the self-fulfilling prophecy are all expressions of this dual commitment: specific enough to be empirically grounded, structural enough to explain patterns that transcend individual intention.
The Self-Fulfilling Prophecy. A false definition of a situation evokes behavior that makes the originally false conception come true. The self-fulfilling prophecy is self-concealing: the outcome appears to validate the belief, and the validation discourages examination of whether the belief was accurate in the first place. In the AI transition, the prophecy of obsolescence and the prophecy of enduring value are both self-fulfilling—the outcome is determined not by the technology but by which definition structures institutional behavior. Interrupting the negative prophecy requires the structural equivalent of deposit insurance: credible institutional commitment to a different definition of the situation.
The Thomas Theorem. “If men define situations as real, they are real in their consequences.” W. I. Thomas’s theorem, which Merton built upon, insists that subjective definitions of situations have objective consequences regardless of their accuracy. In the AI displacement discourse, multiple competing definitions are circulating simultaneously: the triumphalist definition (transformative opportunity), the elegist definition (cultural loss), and the numbed silence of those who feel both and articulate neither. Each definition evokes different behaviors, and different behaviors produce different material realities. The discourse is not commentary on the transition. It is an input into the transition’s outcome.
Manifest and Latent Functions. Every institutional practice serves stated purposes (manifest) and unstated ones (latent). Manifest and latent functions diverge most damagingly when the latent functions undermine the conditions required for the manifest functions to succeed. AI adoption with the manifest function of productivity improvement and the latent function of dependency reduction may eliminate the mentorship conditions under which the judgment that AI requires cannot develop. The rain dance does not produce rain. But if it destroys the social solidarity that kept the community together during the drought, the failure is worse than merely meteorological.
The Matthew Effect. Advantage accumulates through feedback loops between position and returns. The Matthew Effect in the AI transition operates through at least four channels: the capability channel (AI amplifies existing expertise, widening the gap between the highly skilled and the less skilled); the institutional channel (established companies convert AI-augmented output into market value faster than startups); the network channel (dense professional networks accelerate the translation of capability into opportunity); and the cognitive channel (effective AI use requires prior cognitive development that is itself unequally distributed). Moderation requires structural intervention in all four channels, not just access to the tool.
Unanticipated Consequences. Merton identified five structural sources of unintended consequences in his 1936 essay: ignorance, error, the imperious immediacy of interest, values, and self-defeating predictions. Unanticipated consequences in the AI transition include the intensification of work without deepening of engagement (Berkeley 2026 study), the erosion of human mentorship as a pathway to expertise, and the self-concealing feedback loops by which withdrawal from professional investment produces the diminished professional value that justified the withdrawal.