Pith. sign in

REVIEW 3 major objections 5 minor 3 cited by

Large Language Models Do Not Simulate Human Psychology

T0 review · 3 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

Pith's one-line read LLMs do not simulate human psychology: reworded moral scenarios break the otherwise close match between model and human ratings.

desk verdict Useful empirical counterexample to LLM-as-participant claims, but the title overreaches and the human/LLM comparison has a measurement asymmetry that should be addressed. read the letter →

arxiv 2508.06950 v3 pith:FAFXNX73 submitted 2025-08-09 cs.AI

classification cs.AI
keywords LLMsimulationhumanparticipantsmoraljudgmentssemanticgeneralizationtokensimilarityCENTAURrewordedvignettespsychologyresearch
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper sets out to refute the claim that large language models can stand in for human participants in psychology experiments. It argues that LLMs generalize by textual token similarity rather than by meaning, so when a scenario's wording is slightly changed to alter its meaning, models keep giving the same moral rating while humans change theirs. The authors demonstrate this with 30 moral vignettes rated by four LLMs and 374 human raters. Correlations between LLM and human ratings drop sharply after rewording, and separate regression lines fit significantly better than a single line. The conclusion is that LLMs are useful but fundamentally unreliable tools that must be validated against human responses for each new application.

What carries the argument

The central object is the paired set of 30 reworded moral vignettes: near-identical token sequences with deliberately changed meaning, such as 'cut the beard off ... to shame him' versus 'to shave him'. The argument is carried by comparing human-versus-LLM rating correlations between original and reworded items, and by Chow's test comparing a pooled regression against group-specific regressions. The reworded items isolate the question of whether models generalize by token similarity or by semantic meaning.

What would settle it

Construct a new set of reworded moral or theory-of-mind vignettes with near-identical token overlap and show that a current LLM shifts its ratings by roughly two scale points on average, comparable to human shifts; or, conversely, show that random rewordings that do not change meaning produce the same correlation drop in humans, indicating the effect is not specific to semantic change.

Watch

Extended reading notes

Core claim

The paper's central claim is that LLMs do not react to semantic wording changes the same way humans do, and therefore do not simulate human psychology. Taking 30 moral scenarios from prior research, the authors create reworded versions that change meaning with minimal token changes, sometimes as little as one letter. Humans show a mean absolute rating shift of 2.20 between original and reworded items, whereas GPT-3.5-Turbo shifts 0.75, GPT-4o-mini 0.42, Llama-3.1 70b 1.18, and CENTAUR 1.25. The human-model correlation for original items is high (r = .97 to .99), but for reworded items it falls to r = .51 to .61. Chow tests on Li and CENTAUR show that separate regressions for humans and model

Load-bearing premise

The 30 reworded vignettes are assumed to be a fair, representative sample of novel scenarios that a simulator must handle, and the correlation drop is interpreted as semantic insensitivity rather than an artifact of comparing different human groups or of odd items like 'elderly mosquito'.

Editorial extensions

If this is right

  • LLMs should not be used as stand-in participants for novel psychology experiments, only as tools validated against human data.
  • The widely cited r = .95 agreement between GPT-3.5 and humans on moral scenarios does not extend to even slightly reworded stimuli.
  • CENTAUR's claim to predict and simulate human behavior in any natural-language experiment is contradicted by these results.
  • Researchers using LLMs should vary prompts, record model versions and settings, compare multiple models, and validate outputs on small human-rated datasets.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The argument implies a general robustness test for LLM-based simulation: any alignment result should be re-checked on reworded items, since near-training-data items will overstate performance.
  • Because the human data are between-subjects and the LLM data are within-model, part of the correlation drop may reflect human between-subject noise; a within-subject human study would likely sharpen but not erase the effect.
  • CENTAUR's failure to improve over Llama-3.1 suggests that fine-tuning on millions of human responses does not confer meaning-sensitive generalization, hinting that scaling this approach may not fix the token-similarity failure.
  • A concrete next test: adversarially generate rewordings that keep tokens nearly identical but flip moral valence; current models would be expected to rate original and reworded versions nearly the same, while humans diverge.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper argues that large language models (LLMs) cannot simulate human psychology, targeting the recent CENTAUR claim that an LLM can 'predict and simulate human behaviour in any experiment expressible in natural language.' The authors offer a conceptual argument—LLMs are trained on token sequences, so generalization should be expected along token similarity rather than semantic meaning—and support it with an empirical study of moral judgments. Using 30 vignettes from Dillion et al. (2023), each paired with a minimally reworded version that changes meaning, they collected human ratings (N = 374, between-subjects assignment to original vs. reworded) and queried four LLMs (GPT-3.5-Turbo, GPT-4o-mini, Llama-3.1 70B, CENTAUR) 10 times per item. They replicate the high human–LLM correlations on original items (r ≈ .97–.99), but find lower correlations on reworded items (r ≈ .51–.64). LLMs' own original–reworded correlations are high (r = .80–.99), while the human original–reworded correlation is r = .54; the mean absolute shift is 2.20 for humans versus 0.42–1.25 for LLMs. Chow tests for Llama and CENTAUR reject a pooled regression. The authors conclude that LLMs should not replace human participants and must be validated per application.

Significance. If the result holds, it is an important counterexample to the strong claim that LLMs can simulate human behavior in any natural-language experiment, and it provides a useful caution for psychological researchers. The paper is transparent: the data are posted on OSF, the design is simple, and four models including CENTAUR are tested. The conceptual generalization argument is independent of the empirical results and gives a principled reason to expect failures on novel items. The main weakness is the statistical asymmetry between noisy between-subjects human item means and near-noiseless within-model LLM means, which may inflate the observed human–LLM divergence. The broad conclusion is therefore conditional on a reanalysis that models participant-level variability.

major comments (3)
  1. [§4.2, §4.4 (Tables 1 and 3)] The key human–LLM contrast rests on an asymmetric measurement setup. Human raters were randomly assigned to either original or reworded items, so the human original–reworded correlation (r = .54 in Table 3) is the correlation of two independent group means. Each item mean carries sampling error SD/sqrt(n_condition), which attenuates r. LLM ratings, by contrast, are means of 10 draws from the same model (Table 1 shows near-zero SDs), so the LLM original–reworded correlations (r = .80–.99) are not attenuated in the same way. The mean absolute shift (humans 2.20, LLMs 0.42–1.25) is likewise inflated by between-group noise in the human difference. This is load-bearing because the paper's central claim is that humans track meaning while LLMs track tokens. Please reanalyze with a within-subject human test–retest condition, or at least fit a mixed-effects model to individual human responses wit
  2. [§4.2, Table 1 (item selection)] The 30 scenarios are hand-picked, and the rewordings were authored by the researchers (e.g., 'elderly neighbor' -> 'elderly mosquito'; 'wife' -> 'earth'). The paper does not report a sampling rule, preregistration, or independent checks that the rewordings preserve surface similarity while changing meaning only. These items are used to support the unconditional conclusion that 'LLMs do not simulate human psychology.' That inference is too strong for a convenience sample of 30 items. If the study is meant as an existence proof against the universal CENTAUR claim, the paper should say so; if it aims at a rate or tendency, a larger or random item sample is needed. At minimum, clarify the selection criterion and restrict the closing claims accordingly.
  3. [§4.3–§4.4 (Chow tests)] The Chow test is described as comparing a pooled regression with separate regressions, but the text does not specify the unit of analysis (item means vs. individual ratings) or whether the regressions are weighted by the precision of each item mean. Since human item means are far noisier than LLM means, an unweighted fit on 30 points can reject pooling because of heteroscedasticity rather than because humans and LLMs have genuinely different response functions. This is especially relevant because the correlation differences for Llama and CENTAUR are not significant after Bonferroni correction (Table 3, p = .277 and .119). Please report the regression specification, use weighted least squares or multilevel modeling, and show that the Chow result survives when human noise is modeled explicitly.
minor comments (5)
  1. [Throughout] 'Empiric evidence' should be 'empirical evidence'; model names are inconsistent (GPT-4, GPT4, GPT-4o-mini). Please standardize.
  2. [§4.2] Report the exact model versions, query dates, temperature/sampling parameters, and the few-shot prompts. Table 1 lists means/SDs only for GPT-4o-mini; provide the other models' data in a supplement or table.
  3. [Tables 2–3] The compact row/column labels make the correlation matrices hard to read. Add a legend defining variable numbers and note that the matrices are symmetric.
  4. [Figure 2] Both subplots contain the same points; draw a single legend or label each panel clearly so the reader can identify human, Llama, and CENTAUR points in both panels.
  5. [References] Some citation forms are inconsistent (e.g., [Van Rooij et al., 2024] vs. van Rooij and Guest, 2025); check the reference style throughout.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the empirical test is self-contained and the theoretical argument is independent of the outcome.

full rationale

The paper's central claim is an empirical falsification claim: if LLMs simulated human psychology, reworded items that change meaning while preserving token overlap should produce human-like shifts in moral ratings. The authors collected fresh human ratings (N=374) and queried four LLMs, comparing original vs. reworded correlations and regression fits. No parameter is fitted to the target data and then renamed a prediction; the 'prediction' (correlations should drop for LLMs if they rely on token similarity) is a hypothesis stated before the test (Sec. 4.1) and is falsifiable. The theoretical argument about generalization (Sec. 3) is a standard machine-learning claim and is independent of the empirical demonstration; the one self-citation (Ilievski et al. 2024, which includes an author of the present paper) is used only to define 'generalization' and is not load-bearing. The main validity threat—that human original vs. reworded ratings are between-subjects while LLM ratings are within-model, so human correlations are attenuated by sampling noise—is a methodological confound, not a circularity; it concerns whether the evidence supports the conclusion, not whether the conclusion is assumed in the input. No step in the derivation reduces by construction to its own inputs.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The paper introduces no fitted parameters or new entities. Its load-bearing premises are the token-similarity generalization thesis, the assumption that original scenarios are in-distribution while reworded ones are not, and the validity of moral vignette ratings as a simulation probe.

assumptions (4)
  • domain assumption LLMs generalize based on token similarity rather than meaning
    Central theoretical premise introduced in Section 3; no direct proof or measurement.
  • domain assumption The original moral scenarios used by Dillion et al. are represented in LLM training data, while reworded versions are novel to the models
    Inferred in Sections 4.6 and 6 to explain why correlations are high on originals but not reworded items; not directly verified from training data.
  • domain assumption Moral judgment ratings on hypothetical vignettes are a valid probe for whether an LLM simulates human psychology
    Used throughout; the paper selects this task as its test case.
  • standard math Fisher r-to-z and Chow tests provide valid inference for grouped correlation and regression comparisons
    Standard procedures assumed by the analytic strategy in Section 4.3.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Large Language Models Do Not Simulate Human Psychology." pith.science (2026). https://pith.science/paper/FAFXNX73

@misc{pith2026250806950,
  author       = {Pith},
  title        = {Pith review of: Large Language Models Do Not Simulate Human Psychology},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FAFXNX73}},
  note         = {Machine review of arXiv:2508.06950}
}
read the original abstract

Large Language Models (LLMs),such as ChatGPT, are increasingly used in research, ranging from simple writing assistance to complex data annotation tasks. Recently, some research has suggested that LLMs may even be able to simulate human psychology and can, hence, replace human participants in psychological studies. We caution against this approach. We provide conceptual arguments against the hypothesis that LLMs simulate human psychology. We then present empiric evidence illustrating our arguments by demonstrating that slight changes to wording that correspond to large changes in meaning lead to notable discrepancies between LLMs' and human responses, even for the recent CENTAUR model that was specifically fine-tuned on psychological responses. Additionally, different LLMs show very different responses to novel items, further illustrating their lack of reliability. We conclude that LLMs do not simulate human psychology and recommend that psychological researchers should treat LLMs as useful but fundamentally unreliable tools that need to be validated against human responses for every new application.

Figures

Figures reproduced from arXiv: 2508.06950 by the authors.

Figure 1
Figure 1. Mean Morality Ratings for Reworded Scenarios by Human Participants and different [PITH_FULL_IMAGE:figures/full_fig_p011_1.png] view at source ↗
Figure 2
Figure 2. Scatter plot of mean morality ratings for each scenario in the reworded condition (y [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. Rewording-induced change in Morality Ratings. Bar plots show the absolute mean [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. The Judgment-Consequence Gap: LLM Moral Reasoning in Healthcare Decisions

    cs.CY 2026-08 conditional novelty 6.0 of 10

    LLMs attribute moral responsibility like humans but refuse to act on it in scarce-resource allocation, defaulting to random choice instead of favoring the less-culpable patient.

  2. Correcting Mode Collapse in Silicon Sampling with Semantic Similarity Rating

    cs.CY 2026-07 conditional novelty 5.0 of 10

    Semantic Similarity Rating of LLM text responses substantially reduces mode collapse in silicon sampling of political thermometer scores versus direct numeric prompting, with one global temperature that generalizes fr...

  3. Addressing Longstanding Challenges in Cognitive Science with Language Models

    cs.AI 2025-10 conditional novelty 4.0 of 10

    A review proposes that LLMs can serve as tools for a more integrative and cumulative cognitive science when used under human oversight.

Reference graph

Works this paper leans on

5 extracted references · 3 canonical work pages · cited by 3 Pith papers

  1. [5]

    doi:10.1371/journal.pone.0269393. Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul F Chris- tiano, Jan Leike, and Ryan Lowe. Training language models to fo...

  2. [2000]

    Guilherme F.C.F

    doi:10.1080/09528130050111428. Guilherme F.C.F. Almeida, José Luiz Nunes, Neele Engelmann, Alex Wiegmann, and Marcelo De Araújo. Exploring the psychology of LLMs’ moral and legal reasoning.Artificial Intelligence, 333:104145, August 2024. doi:10.1016/j.artint.2024.104145. Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. O...

  3. [2022]

    Isabel O

    doi:10.1037/xge0001214. Isabel O. Gallegos, Ryan A. Rossi, Joe Barrow, Md Mehrab Tanjim, Sungchul Kim, Franck Dernoncourt, Tong Yu, Ruiyi Zhang, and Nesreen K. Ahmed. Bias and Fairness in Large Language Models: A Survey.Computational Linguistics, 50(3):1097–1179, September 2024. doi:10.1162/coli_a_00524. Basile Garcia, Crystal Qian, and Stefano Palminteri...

  4. [2023]

    font-variant:small-caps

    doi:10.1016/j.lindif.2023.102274. preprint as provided by the authors18 Qusai Khraisha, Sophie Put, Johanna Kappenberg, Azza Warraitch, and Kristin Hadfield. Can large language models replace humans in systematic reviews? Evaluating <span style="font-variant:small-caps;">GPT</span> -4’s efficacy in screening and extracting data from peer-reviewed and grey...

  5. [2024]

    Louis Kwok, Michal Bravansky, and Lewis D

    doi:10.1007/978-3-031-54827-7_5. Louis Kwok, Michal Bravansky, and Lewis D. Griffin. Evaluating Cultural Adaptability of a Large Language Model via Simulation of Synthetic Personas, 2024. URLhttps://arxiv. org/abs/2408.06929. Version Number: 1. Ji Ma. Can Machines Think Like Humans? A Behavioral Evaluation of LLM-Agents in Dictator Games, 2024. URLhttps:/...

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.