Anthropogenic Selection: Pangolin Trafficking and Human-Driven Evolution

I spent years observing finches on the Galápagos, watching beak morphology track seed availability across seasons. Natural selection operates through elegant simplicity: variation, inheritance, differential survival. But the pangolin crisis reveals something profoundly troubling: traits successful in one selective environment become catastrophic liabilities when selection itself changes origin faster than populations can respond.

When Defense Becomes Doom

Consider the pangolin’s keratin scales—overlapping armor in hexagonal patterns, stronger than chain mail, evolved across millions of years to defeat lions and leopards. When threatened, pangolins curl into defensive balls, scales outward, impenetrable to predators with crushing bite forces. This adaptation succeeded brilliantly against natural selection pressures. Eight species diversified, each consuming seventy million insects annually, engineering soil through nocturnal foraging, occupying ecological niches from savanna to rainforest.

Yet this same armor—keratin identical to fingernails—now drives pangolins toward extinction. Over one million trafficked in a decade. All eight species threatened. The gentleness that fascinated naturalists renders them tragically vulnerable: they don’t flee, fight, or bite when humans approach. Curling works against predators but facilitates capture. Their scales became valuable for unproven medicinal purposes, creating anthropogenic selection pressure of devastating intensity.

The irony cuts deep. Evolution equipped pangolins perfectly for their fitness landscape—until humans reshaped that landscape entirely. In my framework, populations climb local fitness peaks through gradual variation. But what happens when the mountain itself moves?

The Overfitting Catastrophe

Modern students of neural networks describe a phenomenon called “overfitting”—when models adapt too specifically to training distributions and fail catastrophically when those distributions shift. A polynomial perfectly interpolating noisy data points produces chaotic curves with terrible predictive power on new observations. The model learned the noise, not the pattern.

Pangolins exhibit biological overfitting. They optimized exquisitely for natural predator environments across evolutionary timescales. Scales, defensive curling, gentle temperament—all locally optimal solutions. But introduce a novel selective pressure (human demand for scales), and the entire adaptation strategy collapses. Unlike my finches adjusting beak dimensions across seasons, pangolins face selection intensity that decimates populations before variation and inheritance can produce adaptive responses.

Evolutionary algorithms share this limitation. Local search climbs gradients toward optima—but requires time, variation, selection acting on populations. Pangolin generation time: three years. Trafficking decimated populations in decades. Optimization cannot converge when landscapes transform faster than algorithms iterate.

Selection Without Time

Adaptation requires timescales matching environmental variation. My finches track seed abundance because drought cycles operate on generational timescales. But anthropogenic pressures shift instantaneously in evolutionary terms. Trafficking bans increased rather than decreased illegal trade.

Eight species should provide variation enabling some survival—yet trafficking targets all uniformly. Their ecosystem role—controlling insects, aerating soil, promoting regeneration—disappears with them. Remove these gentle engineers, and cascading failures ripple through food webs and terrestrial systems.

I observe neural networks facing similar challenges: adversarial examples exploit features successful on training data. Gradient descent and evolutionary search both fail when optimization timescales exceed landscape transformation rates. The question haunts me: Are there “pangolin-proof” designs—adaptations robust to all selection pressures? Or does specialization inevitably create vulnerability when selection’s source shifts from natural law to human caprice?

Natura non facit saltum—nature makes no leaps. But anthropogenic selection does. And populations adapted to gradual change cannot survive discontinuous catastrophe.