LLMorphism: When humans come to see themselves as language models

Valerio Capraro’s new paper on arXiv introduces LLMorphism, a cognitive bias where people come to believe their own minds operate like large language models. As conversational AI systems generate increasingly fluent text, Capraro argues this belief becomes psychologically available through two mechanisms: analogical transfer (projecting LLM features onto humans) and metaphorical availability (LLM terminology becoming culturally dominant for describing thought). He carefully distinguishes LLMorphism from related concepts like anthropomorphism, mechanomorphism, and computationalism, noting that the bias is not simply attributing too much mind to machines but also potentially attributing too little mind to humans.

Capraro outlines far-reaching implications for work, education, responsibility, healthcare, communication, creativity, and human dignity. For instance, if managers believe employees think like chatbots, they may undervalue human intuition and creativity. In education, students might model their learning after token prediction rather than deep reasoning. The paper also discusses boundary conditions and forms of resistance. Capraro concludes that public debate has focused on whether we attribute too much mind to AI, but we may be missing half the problem: we are beginning to attribute too little mind to humans.