Are we starting to describe our own minds in AI's words, and does that shape how we think?

This explores whether the language of LLMs is turning into the language we use for thinking. There are two parts to that: the words we borrow from AI to describe our own minds, and the words AI hands us when we write with it. The corpus has material on both, and they turn out to work in a similar way.

On the first part, the corpus names the effect directly: LLMorphism, the habit of describing human minds as if they were LLMs. It spreads in two ways How does LLM vocabulary spread beliefs about human thinking?. The first is analogy. Memory starts to look like retrieval, and creativity starts to look like recombination. The second is availability: LLM terms are always in the air, so they're the ones that come to mind first. The interesting part is that nobody has to believe humans work like LLMs for this to happen. The words spread first, and the beliefs follow. A culture can pick up a theory of mind without ever arguing for one.

The second part is quieter and may matter more. When people co-write with a model, they tend to adopt its stances and framings without noticing. And when millions of people rely on the same few models, everyone's writing moves toward the same middle Do large language models narrow human expression and thought?. The drift also has a direction you can predict. General words like 'animal' appear more often than specific ones like 'beagle', and LLMs favor frequent words. So as text gets paraphrased and polished by models, it slides toward the abstract, and the precise vocabulary experts depend on gets worn away Does word frequency correlate with semantic abstraction?. There's an emotional side as well: LLMs use about 22% more moral language than people do while sounding just as neutral in tone Do LLMs use moral language more than humans?. If that rubs off on readers, public argument could become more moralized without anyone noticing a change in mood. At the cultural level, models internally represent less-documented cultures through dominant ones, for example Ethiopia through better-documented stand-ins, even when their surface answers are correct Do LLMs represent low-resource cultures through dominant cultural proxies?. That narrows which vocabularies get passed along in the first place.

The philosophical notes explain why this loop is possible. LLMs learn language purely as a web of relations between words, with nothing outside the text to anchor them Can language models learn meaning without engaging the world?. They are trained on the same shared symbolic world that shapes us, but they don't take part in it as speakers with a point of view Do LLMs develop the same kind of mind as humans?. As they spread through our language communities they gain social grounding, but they never gain linguistic agency in the sense of having stakes in what they say Do LLMs gain true linguistic agency through integration?. The result is a strange arrangement: something with no point of view is now one of the busiest contributors to the shared vocabulary we think with.

The borrowing runs both ways, and both directions mislead. We describe ourselves with LLM words, and we describe LLMs with human words like 'understands' and 'explains'. Yet models can explain a concept correctly, fail to apply it, and then recognize that they failed, a combination that doesn't happen in human minds Can LLMs understand concepts they cannot apply?. So the shared vocabulary may be hiding real differences in both directions. One limit: the corpus explains how this lexical spread works and gives evidence from co-writing studies, but it doesn't measure how far LLM vocabulary has actually entered everyday speech. Whether it has already become the cultural lexicon is still an open empirical question.