The Matthew Effect

In the AI Story

Merton's original empirical work documented the effect in scientific publishing, where co-authored papers between a Nobel laureate and an unknown researcher generated citations that accrued almost entirely to the laureate. The junior researcher's equivalent contribution disappeared into the laureate's accumulating reputation. But Merton recognized the pattern was not specific to science—it was a general feature of social systems with network effects, cumulative causation, and positive feedback loops. The pattern has since been documented in economics (wealth concentration), education (achievement gaps that widen over time), technology adoption (platform dominance), and urban development (why successful cities become more successful).

The mechanism operates through multiple channels simultaneously. At the individual level, the researcher who receives early recognition attracts better collaborators, whose contributions enhance the quality of subsequent work, which attracts more recognition. At the organizational level, the company with initial market share attracts more users, which generates more data, which improves the product, which attracts more users. At the national level, the country with infrastructure advantages attracts talent and investment, which builds more infrastructure, which attracts more talent and investment. Each channel follows the same mathematical structure: a variable that is positively correlated with its own rate of change.

The AI transition amplifies the Matthew Effect through a mechanism that is both simple and devastating: AI tools are capability multipliers, and multipliers preserve ratios. If one developer is twice as capable as another, and both receive access to a tool that doubles capability, the absolute gap widens even as both improve. The San Francisco developer with fast internet, professional networks, venture capital access, and institutional support receives more from AI than the Lagos developer with equivalent talent but fewer complementary assets—not because the tool discriminates, but because the tool amplifies whatever the user brings to it, and what the user brings is shaped by the structural environment she inhabits.

Recent scholarship has confirmed the Matthew Effect is intensifying in the AI era rather than diminishing. A 2026 analysis in Network Law Review found that AI-assisted academic publishing concentrates reputational gains among already-established scholars. The researchers with strong foundations use AI to produce more and better work; researchers without those foundations produce more output that receives less recognition. The gap widens, and it widens through a mechanism that appears, on the surface, to be democratizing: equal access to tools produces unequal outcomes when the tools amplify unequal inputs.

Origin

The term appeared in Merton's 1968 essay 'The Matthew Effect in Science,' published in Science magazine. The choice of the Biblical verse was deliberate: Merton wanted a label that would communicate the mechanism's ancient character (this is not a new problem) while providing a vivid image (abundance to the abundant, scarcity to the scarce) that would make the concept memorable. The strategy succeeded—the Matthew Effect became one of the most widely cited concepts in social science, applied far beyond its original domain.

Merton later regretted the term slightly, because the religious allusion led some readers to treat it as a moral principle rather than a structural description. He emphasized repeatedly that the Matthew Effect is not a claim about what should happen—it is a claim about what does happen in systems with certain structural properties. Identifying the effect is not endorsing it; it is the prerequisite for addressing it through institutional intervention.

Key Ideas

Cumulative Causation. Advantage feeds advantage—recognition attracts resources, resources enable productivity, productivity generates recognition—in self-reinforcing cycles that compound over time.

Structural, Not Moral. The effect describes what happens in systems where advantages accumulate; it is not a claim about desert, fairness, or the intrinsic worth of those who benefit or lose.

Operates at Multiple Scales. The same mathematical structure governs individual careers, organizational growth, national development, and civilizational trajectory—making the effect one of sociology's most transferable concepts.

Requires Complementary Assets. AI amplifies capability, but realized value depends on complementary assets (networks, capital, infrastructure) that are distributed far more unequally than tool access.

Intervention Must Be Structural. Individual effort cannot overcome cumulative structural disadvantage—addressing the Matthew Effect requires institutional redistribution of the complementary assets that make tools valuable.