“Because LLMS statistically mimic the language people have used, they often fool people into thinking that they operate like people.”
Our reading, the sentence that was checked:
Because LLMs statistically mimic human language, they often fool people into thinking they operate like people.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Why DO large language models hallucinate?, paragraph 14.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-why-do-large-language-models-hallucinate-30dec2-a1
We recorded 16 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2025-05-05. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Causal claim
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No horizon
- Deadline state on 2026-10-29
- No horizon (the release’s reference date)
From the speaker
- Because LLMS statistically mimic the language people have used
- often fool people
- thinking that they operate like people
Our interpretation
- The ordinary reading is that humanlike language causes recurring mistaken attribution of humanlike mental processes.
- Merely finding prose humanlike, trusting an answer, or mistaking an unidentified chat partner for a human is relevant but does not by itself establish beliefs about a known LLM's internal processes.
- No numerical threshold is imposed on 'often'. The assessment concerns systems and user beliefs existing by the post date.
Counts as supported if: Controlled studies or similarly strong causal evidence show that humanlike LLM language repeatedly increases users' mistaken attribution of humanlike cognitive processes.
Counts as contradicted if: Well-matched evidence across relevant settings shows that humanlike language does not produce the claimed attribution, or that the attributed processes are not mistaken.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Does the evidence support the cause?
Outcome: Unresolved
The record was checked, and the evidence available by the cutoff does not settle it.
Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.
- Evidence as of
- Deadline state when assessed
- No horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
The studies establish related phenomena: presentation affects anthropomorphism, and an LLM can be mistaken for a human in conversation. They do not settle Marcus's more specific explanation that statistical language imitation often makes people believe known LLMs operate through humanlike mental processes. The grammatical-person experiment also found no effect on anthropomorphism. These findings neither establish nor refute the full causal claim.
Limitations
- A broader reading limited to apparent humanlike behavior would receive stronger support.
- The cited studies are preprints.
- Study years identify the reported research; precise data-collection dates were not established.
Evidence (2 items)
Context only: arXiv; Michelle Cohn and colleagues,
no effect of Grammatical Person
Why it matters here: The experiment found greater anthropomorphism with speech plus text, but no grammatical-person effect. It supports a role for presentation cues while limiting the inference that humanlike wording itself explains the claimed belief.
- Publisher
- arXiv; Michelle Cohn and colleagues
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: arXiv; Cameron R. Jones and Benjamin K. Bergen,
a win rate of 73%
Why it matters here: Participants often selected the persona-prompted model as the human conversational partner. This establishes successful imitation in the test, but does not measure whether people who know they are using an LLM believe it has humanlike internal processes.
- Publisher
- arXiv; Cameron R. Jones and Benjamin K. Bergen
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
Studies that manipulate humanlike language while holding other cues fixed, measure explicit beliefs about a known LLM's cognitive processes, and test whether those beliefs are inaccurate.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Why DO large language models hallucinate?
Because LLMS statistically mimic the language people have used, they often fool people into thinking that they operate like people.
- Where
- paragraph 14
- Published
- 2025-05-05
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-why-do-large-language-models-hallucinate-30dec2-a1: Current. Outcome: Unresolved. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.
About assessment marcus-why-do-large-language-models-hallucinate-30dec2-a1.