Even now, somewhere, a person is asking a chatbot to explain a concept they could have looked up themselves, and a second person is asking the same chatbot to stress-test an argument they already understand cold. According to a new peer-reviewed paper in the journal Societies, these are not two people having similar experiences with the same tool. They are, dear reader, becoming two different kinds of citizen.
A Concept With a Name Now
The paper, "AI and the Rise of Societal Bifurcation: Cognitive Dependency, Inequality and Democratic Pressure," comes from Michael Gerlich, and it does what academic papers are supposed to do and rarely manage: it gives an anxiety a name. Societal bifurcation is Gerlich's term for a split that isn't about who has access to AI — everyone increasingly does — but about who keeps their interpretative autonomy intact while using it, and who quietly lets it atrophy. Same tool, opposite trajectories.
The Mechanism: Cognitive Offloading
The engine underneath the split is what the paper calls cognitive offloading — the habit of handing a machine the parts of thinking that used to be yours: evaluating a claim, weighing a source, deciding whether an explanation actually holds together. Do this unstructured, over and over, and the paper finds it produces reduced metacognitive monitoring — the academic way of saying you get worse at noticing when you don't understand something — paired with inflated confidence in answers you're no longer checking.
Set against that is a resilient minority who use the same tools as an amplifier: a second opinion to argue with, not a verdict to accept. Same input, and yet — as if this were not enough — wildly different output, depending on the habits of mind the user brought to the interaction in the first place.
The Part Where It Gets Into Your Paycheck
This isn't confined to the seminar room. The paper ties cognitive dependency directly to labor-market vulnerability: workers who never developed the habit of engaging AI critically are, per the analysis, less adaptable and more exposed to automation than workers who use the same tools as a thinking partner. Worse, it's circular — economic insecurity pushes people toward AI as a cheaper substitute for effort, which erodes the very reflective capacity that would have made them more resilient in the first place. Sources within the labor economics community confirm this is exactly as unkind as it sounds.
Democracy, Meet Synthetic Persuasion
The paper's darkest turn is political. A public with declining interpretative autonomy is, it argues, dramatically easier to move with the synthetic, mass-produced persuasive content generative AI is uniquely good at manufacturing at scale. The two classes here aren't symmetrical: the amplifier minority can spot a manipulated narrative for what it is; the substitute majority increasingly cannot — and that asymmetry is precisely what erodes institutional trust and democratic stability. How long, one wonders, before that gap is the whole ballgame.
Not Fate, Just the Current Trajectory
To its credit, the paper stops short of declaring this destiny. Gerlich frames bifurcation as contingent, not deterministic — shaped by the educational, institutional, and organizational environments that either build reflective AI habits into how people use these tools, or leave everyone to sort it out alone at 11pm with a chatbot and a deadline. Which is, in fairness, most of us. The proposed fix amounts to: teach the amplifier habit before the substitute habit becomes the default setting for an entire generation.
Two classes, one login screen, and dear reader, only one of them knows which class it's in.
