Creeping Normalcy

In the AI Story

The original formulation emerged from Diamond's fieldwork in Montana and his analysis of historical cases where environmental deterioration accumulated across decades or centuries without triggering adequate institutional response. In the Norse Greenland case, the grasslands eroded year by year, but each year's pastureland looked only marginally thinner than the previous year's. The cumulative loss was catastrophic; the incremental loss was unremarkable. The same dynamic operated on Easter Island (tree by tree), among the Maya (field by field), and in the Montana valleys Diamond studied directly.

The cognitive mechanism is specific. Human perception is tuned to changes from a baseline, and the baseline itself adjusts to current conditions. What was striking in year one becomes normal by year five. What would have shocked a visitor from the past is invisible to the resident who has experienced the transition gradually. This produces a systematic bias toward underestimating cumulative change — and the bias operates most powerfully on the very people whose livelihoods depend on accurate perception of environmental conditions.

Applied to the AI transition, creeping normalcy requires a counterintuitive extension. The changes in the cognitive economy between October 2025 and March 2026 were not slow — they were dramatic, visible, and widely discussed. Yet the normalization occurred at a speed that rivaled the change itself. The mechanism is different: not imperceptibility of change, but failure to sustain attention across the timescales institutional adaptation requires. The media cycle metabolized the shock; public attention moved on; institutions continued operating according to frameworks that assumed the pre-shock environment. The outcome — failure to recognize that adaptation is required — is identical to what Diamond documented in slow cases. The mechanism adapted to the speed of contemporary information processing.

A deeper form of the phenomenon — landscape amnesia, Diamond's term from The World Until Yesterday — operates across generations. Each generation perceives its own conditions as the baseline and loses awareness of what came before. Applied to AI, this means that developers who began their careers in the AI-augmented environment have no experiential baseline for pre-AI cognitive work. They cannot miss the depth they did not develop because they do not know what depth feels like. The compound effect across generations produces the invisible depletion that cognitive resource depletion describes.

Origin

Diamond introduced the concept in Collapse (2005), drawing on his Montana fieldwork and his comparative analysis of societies that had failed to perceive cumulative environmental change. The extension into landscape amnesia came in The World Until Yesterday (2012). The Boiling Frog metaphor (which Diamond used as illustrative but acknowledged was biologically imprecise) became the popular shorthand for the concept, though the underlying mechanism is documented in human perceptual psychology independent of the frog.

The application to rapid change — the version relevant to AI — requires modification. Diamond's original formulation assumed slowness was necessary for normalization; the AI case demonstrates that information saturation and attention decay can produce functionally equivalent normalization at high speed. The extension is analytically faithful to Diamond's framework even though he did not make it explicitly.

Key Ideas

Normalization happens below awareness. Each increment of change resets the perceptual baseline, making cumulative change systematically underestimated.

Speed is not a protection. Fast change can be normalized through information saturation and attention decay just as effectively as slow change is normalized through imperceptibility.

Landscape amnesia compounds across generations. Each generation's baseline is the product of prior cumulative change, which it cannot perceive — producing progressive loss across generations without any single generation experiencing the full trajectory.

Institutional memory is the corrective. The societies that resisted creeping normalcy — the Tokugawa with their forest inventories, Icelanders with their commons records — were the ones that institutionalized measurement across timescales longer than any individual's perception.

The AI transition requires new monitoring architecture. Because the change is fast and the normalization faster, contemporary institutions need measurement systems that track what is being lost at the timescales relevant to cognitive development and professional formation — not just quarterly productivity metrics.

Debates & Critiques

Some critics argue that the concept is too flexible to falsify — that any collapse can be attributed to creeping normalcy in retrospect. Defenders have responded that the concept makes specific predictions about institutional vulnerability: societies without long-horizon measurement systems should be systematically more vulnerable to cumulative change than societies with such systems, and the pattern holds across Diamond's paired cases. The extension to fast change is more contested, but the empirical pattern — rapid normalization of dramatic AI developments, measurable by the attention-cycle analysis of media coverage — supports the adapted version of the concept.