Novelli, Floridi, Larsson, Taddeo, and Winter (2025) deploy cognitive linguistics to trace how three metaphorical patterns, "intelligence," "black box," and "hallucinations," structure AI regulatory discourse and generate conceptual path dependence. Their analysis identifies the problem with precision but leaves three questions open: why certain regulatory metaphors persist while competitors fail, what causal mechanism underlies the epistemic deference they rename as "deferential hazards," and what existing legal instruments already address the normative consequences they identify. This paper proposes answers from three frameworks developed independently for analyzing institutional persistence, intentionality mismatch, and memetic evolution. Extended Phenotype Theory (EPT) reconceptualizes regulatory metaphors as cultural replicators whose survival depends on reproductive fitness rather than descriptive accuracy, explaining why "artificial intelligence" outcompetes "stochastic parrots" despite inferior analytical precision. Heteronomous Bayesian Updating (HBU) provides the causal mechanism that "deferential hazards" describes but does not formalize: users calibrate beliefs through interlocutor reactions rather than outcome evaluation, and systems optimized for satisfaction systematically corrupt this process. Asymmetric Intentionality Theory (AIT) formalizes what cognitive linguistics identifies as "personification": users attribute Level 3 intentionality (recursive moral reasoning) to systems operating at Level 1 (preference function optimization), generating self-reinforcing Dynamic Classification Failure. The three frameworks converge on a practical implication: objective liability regimes for risky activities, present across continental civil law systems and functionally paralleled by strict liability doctrine in common law jurisdictions, already provide the legal infrastructure to address design-choice responsibility without requiring new regulatory metaphors or administrative regimes. Argentine civil law (Articles 1757-1758, Civil and Commercial Code) illustrates the point with particular clarity. The convergence of cognitive linguistics and evolutionary jurisprudence on the same phenomena, from independent premises, constitutes evidence for both research programs.
Ignacio Adrián LERER (2026) studied this question.