Governing the Creature We Imagine: Anthropomorphism, LLMorphism, and Institutional Extended Phenotypes in AI Regulation

Public descriptions of artificial intelligence increasingly attribute to technical systems the properties of agents, social groups, civilizations, conspirators, and potentially living subjects. This conceptual paper integrates research on AI narratives, anthropomorphism, LLMorphism, policy agenda-setting, and Extended Phenotype Theory. It proposes that a successful narrative can operate as an institutional extended phenotype when it persists beyond individual speakers, reallocates perceived agency, organizes coalitions, narrows conceivable policy responses, and becomes embodied in durable research, legal, administrative, educational, labor, or market structures. The paper incorporates modest and null findings from lexical-framing experiments, specifies rival explanations and falsifiable hypotheses, and presents a time-indexed causal architecture and empirical research agenda. It does not claim that anthropomorphic language has been shown to cause restrictive regulation or net harm to innovation.