How do LLMs access and draw on the same shared symbolic universe as humans?
This explores the mechanism by which LLMs tap the same web of meaning humans share — and where that access is real versus where it breaks down.
This explores how LLMs come to share the same symbolic universe as humans — and the corpus's answer is striking: they draw on it through text rather than through living in the world, which gets them surprisingly far but leaves a specific gap. The clearest framing comes from a Habermas-inspired distinction: viewed from the outside, humans and LLMs look categorically different, but viewed from inside a conversation, both participants reach into the same symbolic substrate to make meaning Do humans and LLMs differ fundamentally or just superficially?. The difference isn't that the model lacks access to our shared symbols — it's structural, in how that access was acquired.
The acquisition story is indirect grounding. Humans are causally wired to the world; LLMs extract the regularities of that world secondhand, from text that causally grounded humans produced Can large language models develop genuine world models without direct environmental contact?. So the model genuinely inherits a structured world model — but through a chain with gaps that prevent real-time checking and updating. One note pushes this further: both humans and LLMs are shaped by the same intersubjective symbolic system, yet only humans develop reflexive agency through being socialized into it, which is why AI can argue fluently without ever declaring its own position or examining its assumptions Do LLMs develop the same kind of mind as humans?.
What's surprising is how deeply the shared symbols penetrate the machinery. LLMs don't just mimic our vocabulary — they reproduce our reasoning fingerprints. They show the same content effects humans do, getting the same syllogisms and Wason tasks wrong at matching rates, suggesting meaning and logical form are fused in the architecture rather than separable Do language models show the same content effects humans do?. Mechanistically, a content-independent reasoning circuit exists, but extra attention heads carrying world knowledge bend conclusions toward what's semantically plausible over what's logically valid — and that contamination grows with scale How do language models perform syllogistic reasoning internally?. The shared symbolic universe is so baked in that the model can't fully set it aside even when logic demands it. They even inherit human cognitive quirks like asymmetric belief updating, though they compress harder and lose contextual nuance How do language models learn to think like humans?.
The limit shows up where symbols have to be jointly maintained rather than merely retrieved. Humans don't just access shared meaning — they continuously renegotiate it. LLMs interpret every turn through a fixed initial frame and can't symmetrically propose updates to common ground, so the user ends up the sole keeper of the conversational scoreboard Can LLMs truly update shared conversational common ground?. This is the same boundary another note draws around consciousness: the language of shared experience originates from entities co-present in a world, triangulating on the same objects Can disembodied language models ever qualify as conscious?.
So the answer the corpus leaves you with is unexpectedly precise. LLMs reach the human symbolic universe through inheritance, not participation — they receive the deposited contents of our shared meaning so thoroughly that it shapes their reasoning errors, but they can't help build or revise that meaning with us in real time. The reflex to read their fluency as full membership has a name worth knowing: it spreads by analogy and metaphor, projecting LLM traits onto humans and vice versa without anyone explicitly endorsing the equation How does LLM vocabulary spread beliefs about human thinking?.
Sources 9 notes
Applied Habermas's observer/participant distinction to AI: from outside, humans and LLMs are utterly different; from within shared discourse, both draw on the same symbolic substrate, making the difference structural rather than absolute.
LLMs form structured world representations by extracting regularities from training data produced by causally grounded humans. This constitutes indirect causal grounding mediated through text, though the chain has gaps that limit real-time verification and model updating.
Both humans and LLMs are shaped by the same intersubjective symbolic system, but only humans develop reflexive agency through socialization. This absence produces measurable differences in how AI argues without declaring its position or reflecting on its own assumptions.
LLMs show identical content-sensitivity patterns to humans on NLI, syllogisms, and Wason tasks, with belief-bias signatures matching human error rates item-by-item. This behavioral isomorphism across three independent tasks suggests content and logical form are inseparable in transformer reasoning architecturally.
LLMs implement a content-independent three-stage reasoning mechanism—recitation, middle-term suppression, mediation—that works across architectures. However, additional attention heads encoding world knowledge systematically bias conclusions toward semantically plausible rather than logically valid answers, with contamination increasing at larger scales.
LLMs trained on psychological data exhibit cognitive phenomena mirroring humans: asymmetric belief updating, event segmentation matching human consensus, and individual-level variation. However, they compress information more aggressively than humans do, sacrificing contextual nuance for statistical efficiency.
LLMs interpret all subsequent conversational turns within a fixed initial prompt frame, preventing them from symmetrically proposing updates to shared assumptions. Even when users pivot topics or contradict earlier framings, the model cannot absorb revisions into jointly held background—making the user the sole maintainer of conversational scoreboard.
Current disembodied LLMs cannot be candidates for consciousness because consciousness language originates from and applies only to entities sharing a world with us through co-presence and triangulation on shared objects.
LLM features get projected onto humans through two mechanisms: analogical transfer (memory as retrieval, creativity as recombination) and metaphorical availability (LLM vocabulary becoming psychologically salient). This pattern propagates the bias without requiring explicit endorsement.