The Cognitive Obesity Hypothesis: AI and the Risk of Mental Atrophy

For two centuries, the mechanization of physical labor delivered enormous gains in productivity and comfort while quietly eroding the bodies it served. Adults in wealthy countries now grip with roughly 20% less force than their counterparts did in the mid-1980s, expend on the order of 100-140 fewer occupational kilocalories per day than in the 1960s, and live in environments where physical inactivity is implicated in more than five million premature deaths each year. Here we ask whether the rapid spread of generative artificial intelligence and increasingly autonomous AI agents could be starting a structurally similar transition in the cognitive domain, and we advance this as a testable risk hypothesis rather than an established fact. Reading the physical case as a worked example, we argue that the same three-step mechanism that drove bodily decline is now plausibly acting on cognition: substitution of effort, redesign of the surrounding environment, and a rebound in which a cheaper resource is consumed in greater quantity (the Jevons paradox). We identify five early, individually preliminary signals consistent with that hypothesis: (1) reduced neural engagement during AI-assisted writing; (2) cognitive offloading and lower critical-thinking scores among adults who delegate reasoning to chatbots; (3) AI dependence associated with weaker reasoning among students; (4) deskilling in knowledge work, accelerated as passive assistants give way to agents that perform whole tasks; and (5) the displacement of the desirable difficulties known to consolidate learning. We weigh these signals against the strongest opposing case (that cognitive tools augment rather than diminish the mind) and against confounders such as the pre-existing reversal of the Flynn effect (the twentieth-century rise in measured intelligence, now declining in some cohorts for reasons unrelated to AI). Our contribution is threefold: a mechanistic parallel transferred from a fully documented, century-scale physical case; a falsifiable cellular prediction, drawn from the biology of effort, that frictionless cognitive offloading may, like disuse in muscle, produce not merely lag but measurable atrophy; and a call for the population-scale, longitudinal measurement that does not yet exist but is the precondition for testing the hypothesis before, rather than after, the fact.