Double Rootlessness: AI's Cognitive Illusion and Systemic Risk Amplification

The phrase "double rootlessness" entered analytical circulation quietly, attached to a February 9th piece that has since drifted through technology criticism at the speed of ideas that are uncomfortable rather than wrong. The argument is architectural. It posits two simultaneous disconnections in AI systems that compound each other in ways that are individually manageable but together become structurally dangerous. The first disconnection is cognitive: these systems produce outputs that pattern-match to understanding without being grounded in anything resembling genuine comprehension. This is not a new observation — the "stochastic parrot" critique made a version of it years ago. What the February piece adds is the second layer. The second disconnection is systemic: as AI becomes embedded in critical infrastructure, financial systems, and decision pipelines, the cognitive illusion at the individual output level gets amplified into something that looks like institutional knowledge but carries none of the error-correction mechanisms that real institutional knowledge develops over time. One rootlessness is a quirk. Two rootlessnesses, operating in parallel, is a load-bearing wall that was never actually installed.

If confirmed, the implications are less about any single AI failure and more about the failure mode we are not watching for. Individual AI errors are increasingly understood and partially mitigated — hallucinations get flagged, outputs get reviewed. What this argument targets is the category of risk that emerges precisely because each individual output seems plausible enough to pass, while the aggregate of those outputs shapes decisions at scale without anyone holding accountability for the systemic drift. The organisations most exposed would be those that have integrated AI into medium-stakes decisions — not the headline applications under scrutiny, but the thousands of quieter deployments in logistics, hiring, content moderation, and risk assessment where no single output is consequential enough to audit but the cumulative directional pressure is enormous. Regulators focused on AI explainability at the individual output level may be solving for the wrong problem entirely.

Watch for any academic or institutional risk literature that picks up the "double rootlessness" framing specifically — if this concept migrates from criticism into safety research, that is the signal that it has found traction where it would matter most.