LLMorphism: When humans come to see themselves as language models

[No public URL — single-author preprint by Valerio Capraro]

Psychology Chatbots Conversation Social Theory Society Cognitive Models Latent

LLMorphism is the biased belief that human cognition works like a large language model. I argue that the rise of conversational LLMs may make this bias increasingly psychologically available. When artificial systems produce human-like language, people may draw a reverse inference: if LLMs can speak like humans, perhaps humans think like LLMs. This inference is biased because similarity at the level of linguistic output does not imply similarity in cognitive architecture. Yet, LLMorphism may spread through two mechanisms: analogical transfer, whereby features of LLMs are projected onto humans, and metaphorical availability, whereby LLM vocabulary becomes a culturally salient vocabulary for describing thought. I distinguish LLMorphism from mechanomorphism, anthropomorphism, computationalism, dehumanization, objectification, and predictive-processing theories of mind. I outline its implications for work, education, responsibility, healthcare, communication, creativity, and human dignity, while also discussing boundary conditions and forms of resistance. I conclude that the public debate may be missing half of the problem: the issue is not only whether we are attributing too much mind to machines, but also whether we are beginning to attribute too little mind to humans.

For most of human history, the production of open-ended, context-sensitive, meaningful language was overwhelmingly associated with human speakers. It is therefore plausible that humans rely on a powerful heuristic: when an entity produces fluent and responsive language, we treat it as though there were a mind behind the words. The overapplication of this heuristic lies at the basis of the anthropomorphization of LLMs. As LLMs become widespread, a reverse inference becomes psychologically available: instead of merely asking whether LLMs are becoming human-like, people may begin to ask whether humans themselves are different from LLMs. This reverse inference may be adopted through at least two mutually reinforcing mechanisms: analogical transfer and metaphorical availability.

These two mechanisms may therefore lead to the reinterpretation of human cognition through the lens of LLMs. This is the mechanomorphism step. It does not consist simply in noticing that LLMs produce fluent language. It consists in using this fact to reinterpret human beings. The first move is justified: LLMs can produce language that resembles human language at the level of observable output. The second move is more questionable: because language is one of the main ways in which humans make thought socially visible, observers may treat similarity in language as evidence of similarity in cognition. This inference, although tempting, is not logically valid, because similarity of linguistic outputs does not entail similarity of underlying cognitive processes. Human language is primarily a tool for communication developed by embodied organisms with needs, emotions, memories, social obligations, developmental histories, and vulnerability to consequences. LLM output, by contrast, is generated by systems trained to model statistical regularities in language, rather than by embodied agents who use language from a situated perspective.

Clearly, these are hypotheses: some may turn out to be false, others may need refinement, and still others may be discovered. The broader point, however, is that public debate on AI has focused mainly on anthropomorphism: whether we are giving too much mind to machines. LLMorphism suggests that this is only half of the problem. The other half is whether we are beginning to take too much mind, agency, and grounding away from humans.