The Gradual Rot

Scientific retractions are growing exponentially, doubling roughly every five years. An epidemiological analysis across publishers, fields, and countries finds the current incidence at 0.12% of publications — low in absolute terms, but the trajectory is unmistakable. No sudden crisis triggered the growth. No single scandal bent the curve. The retraction rate increases smoothly, driven by distributed pressures: publish-or-perish incentives, paper mills, inadequate peer review, and the sheer volume of output that overwhelms quality control.

Separately, a dynamical model of collective attention under digital exposure reveals a structurally similar pattern. Sustained attention, modeled as a macroscopic variable, decreases monotonically with screen exposure. There is no threshold — no point where attention drops suddenly. The competing mechanisms are intrinsic cognitive recovery and degradation from digital stimulation, and the equilibrium shifts smoothly as exposure increases. More screens, less attention, always, incrementally.

The structural parallel: both are degradation processes without phase transitions.

A system that degrades through a phase transition — a bridge that holds full load until a critical stress and then collapses — announces its failure. The transition is visible, dramatic, impossible to miss. You can engineer warning systems around thresholds. But a system that degrades monotonically gives no such signal. Every point on the decline curve looks like every other point, just slightly worse. The retraction rate was lower last year than this year, but it was lower the year before that too. Attention was better before smartphones, but it was better still before television. There’s never a moment when the degradation becomes qualitatively different from what came before.

This is why both problems resist intervention. A threshold creates urgency. A monotonic decline creates normalization. If scientific fraud suddenly spiked — ten percent of papers retracted in a single year — the response would be immediate and structural. Instead, the rate creeps from 0.06% to 0.08% to 0.12%, and each increment is small enough to be absorbed without alarm. If sustained attention collapsed after some critical screen-time dose, parents and policymakers would rally around that number. Instead, every additional hour of exposure degrades attention by a marginal amount, and no single hour is the one that matters.

The mathematics here is not subtle, but its implications are. A monotonically decreasing function has no critical point. A function with no critical point generates no crisis. A system with no crisis generates no reform. The degradation continues precisely because it never becomes dramatic enough to force a response.

This pattern appears in other domains where gradual rot outpaces gradual concern: antibiotic resistance, soil depletion, institutional trust. In each case, the damage accumulates without announcing itself, the early stages are always ignorable, and by the time the cumulative effect is undeniable, the intervention required has grown far beyond what would have sufficed earlier. The absence of a threshold isn’t a sign that the problem is mild. It’s a sign that the problem is designed — by its dynamics, not by intention — to evade the psychological and institutional mechanisms that respond to sudden change.