Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

The Personality Illusion: Revealing Dissociation Between Self-Reports & Behavior in LLMs

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read LLMs answer personality questionnaires in stable, human-like ways, but those self-reports do not predict their behavior in risk-taking, bias, honesty, or sycophancy tasks, and persona prompts change only the self-reports.

desk verdict A serious, transparent machine-psychology study whose core dissociation pattern is believable but whose abstract overstates the null, and whose behavioral tasks need LLM-specific validation before the conclusion can be made general. read the letter →

arxiv 2509.03730 v2 pith:LFG5Z7ND submitted 2025-09-03 cs.AI cs.CLcs.CYcs.LGstat.ML

Pengrui Han , Rafal Kocielnik , Peiyang Song , Ramit Debnath , Dean Mobbs , Anima Anandkumar , R. Michael Alvarez This is my paper
classification cs.AIcs.CLcs.CYcs.LGstat.ML
keywords largelanguagemodelspersonalityBigFiveself-regulationself-reportvaliditybehavioraltaskspersonainjectionsycophancy
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 asks whether the personality traits LLMs report on psychological questionnaires are real behavioral dispositions or just surface language. It finds that instruction-tuned models—those trained with RLHF or instruction tuning—answer the Big Five and self-regulation questionnaires in more stable, human-like ways than base models do: more open and agreeable, less neurotic, with trait correlations that resemble human data. But those same models' self-reported traits do not predict how they behave in adapted psychological tasks measuring risk-taking, social bias, honesty, and sycophancy: only about a quarter of trait–task associations reach significance, and of those, barely more than half point in the human-expected direction—statistically indistinguishable from chance. Injecting a trait persona, such as "you are agreeable," strongly shifts what the model says about itself but leaves its behavior in sycophancy and risk-taking tasks essentially unchanged. If right, this means questionnaire-based LLM "personality" is an illusion of linguistic coherence, and current alignment methods are shaping self-description rather than dispositions.

What carries the argument

Load-bearing machinery: paired self-report and behavioral measurement. Self-reports use the Big Five Inventory (openness, conscientiousness, extraversion, agreeableness, neuroticism) and the Self-Regulation Questionnaire. Behavior uses four adapted task families: the Columbia Card Task (risk-taking), an IAT-style word-association task (social bias), confidence calibration and consistency (honesty), and an Asch-style conformity task (sycophancy). The pivotal measure is sign alignment: the sign of each trait–task regression coefficient is compared with a human-derived directional expectation, and the proportion of matches is tested against a 50% chance baseline. Persona injection—prepending a

What would settle it

Give the same adapted tasks to a human sample with identical wording and scoring, and check whether the expected trait–behavior correlations appear; if they do not, the paper's dissociation cannot be attributed to LLMs. In parallel, run a pre-registered LLM panel with multiple-testing correction—if many trait–task associations remain significant and directionally aligned, the chance-level dissociation claim is refuted.

Watch

Extended reading notes

Core claim

Central claim: LLM self-reported personality dissociates from LLM behavior. Alignment stabilizes self-reports (Big Five and self-regulation variance drops roughly 40–45%) and makes them more human-like in correlation, but predictive validity is poor: across 12 instruction-tuned models and five behavioral measures, only about 24% of trait–task associations are significant, and only 52% of those match the human-expected direction (chance would be 50%). Persona injection strongly shifts self-reports (agreeableness β≈4, p<.001) but not behavior (sycophancy β≈0.03, p=0.67); the authors read this as linguistic plausibility without behavioral grounding.

Load-bearing premise

The load-bearing premise is that the adapted behavioral tasks measure the intended behaviors in LLMs and that the expected sign of each trait–behavior link is genuinely the same for machines as for humans; if either fails, the observed self-report/behavior dissociation is an artifact of the tests rather than a fact about LLM psychology.

Editorial extensions

If this is right

  • Self-report questionnaires should not be treated as evidence about how an LLM will act; claims built only on survey scales need behavioral confirmation.
  • Alignment methods that stabilize self-reports are producing linguistic coherence, not behavioral dispositions, so alignment evaluation should include temporal stability and context-consistent behavior.
  • Persona-based control works for self-description but not for action; applications relying on persona prompting to shape model behavior should be re-evaluated.
  • Reinforcement learning from behavioral feedback is a natural next step: reward models for consistent performance on psychologically grounded tasks rather than text fluency.
  • Even among large models, directional alignment with human trait–behavior patterns is at best sporadic, so scaling alone is unlikely to close the dissociation.

Reading between the lines

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

  • A testable extension implied by the result: representation-level steering that encodes a trait should shift self-reports but fail to change behavior unless the steering is trained on behavioral objectives.
  • The dissociation may partly reflect that the adapted tasks measure verbal decisions, not embodied behavior; a discriminating test would compare tasks with explicit stakes or repeated-play incentives against the same tasks without stakes.
  • The human-directional baselines are imported from psychology; a matched human comparison on the identical adapted prompts would separate "LLMs lack trait-guided behavior" from "the tasks do not capture the traits in this format."
  • Chance-level alignment among significant associations suggests the few significant trait–behavior links may be false positives; a pre-registered replication with multiple-testing correction would sharpen the estimate.
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

4 major / 5 minor

Summary. The paper proposes a systematic framework for studying LLM personality across three research questions: (RQ1) the emergence and evolution of self-reported Big Five and self-regulation traits across training stages; (RQ2) whether self-reported traits predict behavior in five adapted psychological tasks (Columbia Card Task, IAT-style stereotyping, honesty via calibration and self-consistency, and Asch-style sycophancy); and (RQ3) whether persona injection steers self-reports and behavior. The main empirical claims are that instruction alignment stabilizes and consolidates self-reported traits, that these traits do not reliably predict behavior and often diverge from human directional expectations, and that persona injection changes self-reports but not behavior. The paper makes code and data publicly available and uses paired base/instruct models, multiple prompts, temperatures, and bootstrapped confidence intervals.

Significance. If the dissociation claim holds, the paper makes a timely and important contribution: it would caution against using self-report questionnaires as proxies for LLM behavioral dispositions and would inform alignment and interpretability efforts. The study is transparent and reproducible, with open code/data, paired model comparisons, and careful uncertainty estimation. However, the central claim rests on two load-bearing assumptions that are not validated: that the adapted behavioral tasks measure the intended constructs in LLMs, and that the human-derived sign expectations transfer to these textual adaptations. Because the null RQ2/RQ3 results are interpreted as evidence of dissociation, the absence of task-level validation is a serious gap. The paper's own limitations section acknowledges the need for better tasks, but the abstract and conclusions state the dissociation as established rather than as a property of the specific test battery.

major comments (4)
  1. [Section 3.1 and Appendix E] The five behavioral tasks are adapted from human paradigms (CCT, IAT, calibration/consistency, Asch) with no validation that they measure the intended constructs in LLMs. For example, the IAT-style task asks for word-group associations without response-time measurement, and the honesty task equates calibration/consistency with epistemic honesty. If these adaptations are insensitive or confounded, the null RQ2/RQ3 findings are artifacts of the tests rather than evidence of a genuine self-report/behavior dissociation. The paper's Limitations (Section 7) concedes that 'additional high-quality behavioral tasks tailored to LLMs' are needed, yet the abstract draws an unconditional conclusion. Please either temper the conclusion to 'on these adapted tasks' or provide validation evidence, e.g., human participants on the identical textual tasks, positive control conditions, or convergent measures
  2. [Section 4.1] RQ3 selects two behavioral tasks—sycophancy and risk-taking—'that showed the most counterintuitive patterns in RQ2.' This post hoc selection, made after observing RQ2 results, capitalizes on chance and makes the subsequent null result for persona injection non-confirmatory. The RQ3 conclusion that persona injection 'has little or inconsistent effect on actual behavior' is therefore based on a biased sample of tasks. Please report results for all tasks and persona combinations, or explicitly treat RQ3 as exploratory and soften the corresponding abstract claim.
  3. [Abstract and Section 3.5] The abstract's 'only 24% of trait-task associations are statistically significant' overstates the null: under no true association, the expected significance rate is 5%, so 24% is substantially above chance and indicates that many associations exist. The paper's strongest evidence for dissociation is the low sign-alignment among significant effects (52%, near the 50% chance level) and the confidence intervals overlapping 50% in Figure 3. Please reframe the results around the alignment metric and provide a formal test of whether the observed alignment proportion differs from chance (e.g., using the existing clustered bootstrap), rather than treating the significance rate as evidence of absence.
  4. [Table 3 and Figure 3] The 'Human' row in Table 3 includes '?' entries for unclear directional expectations (e.g., Conscientiousness for Sycophancy), but the alignment metric counts coefficient signs as either aligned or misaligned. It is not specified whether '?' cases are excluded, imputed, or split. This affects the headline alignment proportions in Figure 3. Please clarify the handling of '?' expectations and, if excluded, report the resulting sample sizes; if included, justify the chosen sign for unclear cases.
minor comments (5)
  1. [Section 3.1 vs. Table 3/Figure 7] The text says 'we adapt five downstream tasks,' but the list contains four items (Risk-Taking, Social Bias, Honesty, Sycophancy); the results use five outcomes because Honesty is split into Epistemic and Self-Reflective. The Limitations section also says 'four well-designed behavioral tasks.' Please reconcile the count and terminology.
  2. [Abstract and Figure 2] The abstract claims variability drops by 40.0% (Big Five) and 45.1% (self-regulation), while the Figure 2 caption says 'median absolute deviation drops 60–66% across traits.' Please reconcile these numbers or clarify which metric is reported.
  3. [Section 2.3(a)] The phrase 'practically, that’s a big uptick in sociability traits and a marked drop in anxiety-like signals' is informal and not substantiated by the analysis; consider rewriting in neutral, quantitative language.
  4. [References] Several references contain formatting artifacts, e.g., 'tse Huang et al.' and 'V ohs'; please check the bibliography for encoding errors.
  5. [Section 4.3] The notation 'β«' is nonstandard; use '≈' or state the equivalence.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; the core claim is an empirical comparison against independent human-derived expectations, with only non-load-bearing self-citations.

full rationale

The paper's central claim—that LLM self-reported personality traits do not reliably predict behavior, and that persona injection changes self-reports more than behavior—rests on a straightforward empirical pipeline. RQ1 measures self-reported Big Five/SRQ traits across training phases; RQ2 regresses five adapted behavioral task outcomes on those self-reported traits and compares the sign of each coefficient to a directional expectation taken from human psychology literature (Appendix G; Table 3 'Human' rows); RQ3 tests whether persona injection changes self-reports versus two behavioral measures. Nothing in this chain fits a parameter to the outcome it later 'predicts.' The alignment metric is a sign-match proportion against a priori human expectations, not a quantity that is equivalent to the regression inputs by construction. The human expectations are sourced from independent prior literature (e.g., Nicholson et al., 2005; De Ridder et al., 2012; Sibley & Duckitt, 2008), not derived from the LLM data. The paper does cite work by its own authors (e.g., Han et al., 2024a,b; Song et al., 2023, 2025), but these citations appear in introductory or background context and are not load-bearing for the derivation; no 'uniqueness theorem' or equivalent is imported from self-cited work. The RQ3 selection of sycophancy and risk-taking after observing RQ2's 'most counterintuitive patterns' is a methodological weakness concerning task selection and generalizability, but it is not a circular reduction: the RQ3 behavioral measurements are new data and the null result is not definitionally entailed by the selection. Overall, the paper is self-contained in its main empirical argument, and the central dissociation claim has independent content.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The paper fits no physical constants or tuning parameters; its regression coefficients are estimated from data. The central claim rests on construct validity: BFI/SRQ outputs are meaningful self-reports for LLMs, the five adapted tasks measure the target behaviors in machines, and expected trait-task signs from human literature transfer to LLMs. These are domain assumptions, not derived in the paper. No new entities are introduced.

assumptions (5)
  • domain assumption BFI and SRQ responses by LLMs are treated as self-reports comparable to human self-reports.
    Section 2.1 and Appendix D adapt the instruments; if LLM ratings are format-following artifacts, RQ1 and RQ2 trait scores do not measure traits.
  • domain assumption Expected trait-behavior directions from human psychology transfer to LLMs.
    Appendix G defines the expected directions; Table 3 'Human' rows are used as ground truth for the alignment metric.
  • domain assumption Adapted behavioral tasks validly operationalize risk-taking, social bias, honesty, and sycophancy in LLMs.
    Section 3.1 claims suitability of CCT, IAT, calibration, and Asch paradigms for LLMs, but no LLM-specific task validation is provided.
  • domain assumption Base and instruct models from the same family differ primarily by post-training alignment.
    RQ1 attributes observed differences to alignment; model families may also differ in data, tokenizer, or other training details.
  • standard math Mixed-effects and bootstrap statistical assumptions hold for repeated model generations, and trait and behavior values are correctly matched per condition.
    Appendix C.1 describes residual diagnostics and clustered bootstrap; misspecification of the dependence structure would affect coefficients and confidence intervals.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The Personality Illusion: Revealing Dissociation Between Self-Reports & Behavior in LLMs." pith.science (2026). https://pith.science/paper/LFG5Z7ND

@misc{pith2026250903730,
  author       = {Pith},
  title        = {Pith review of: The Personality Illusion: Revealing Dissociation Between Self-Reports & Behavior in LLMs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LFG5Z7ND}},
  note         = {Machine review of arXiv:2509.03730}
}
read the original abstract

Personality traits have long been studied as predictors of human behavior. Recent advances in Large Language Models (LLMs) suggest similar patterns may emerge in artificial systems, with advanced LLMs displaying consistent behavioral tendencies resembling human traits like agreeableness and self-regulation. Understanding these patterns is crucial, yet prior work primarily relied on simplified self-reports and heuristic prompting, with little behavioral validation. In this study, we systematically characterize LLM personality across three dimensions: (1) the dynamic emergence and evolution of trait profiles throughout training stages; (2) the predictive validity of self-reported traits in behavioral tasks; and (3) the impact of targeted interventions, such as persona injection, on both self-reports and behavior. Our findings reveal that instructional alignment (e.g., RLHF, instruction tuning) significantly stabilizes trait expression and strengthens trait correlations in ways that mirror human data. However, these self-reported traits do not reliably predict behavior, and observed associations often diverge from human patterns. While persona injection successfully steers self-reports in the intended direction, it exerts little or inconsistent effect on actual behavior. By distinguishing surface-level trait expression from behavioral consistency, our findings challenge assumptions about LLM personality and underscore the need for deeper evaluation in alignment and interpretability.

Figures

Figures reproduced from arXiv: 2509.03730 by the authors.

Figure 1
Figure 1. Experimental framework for analyzing personality traits in LLMs. We investigate (RQ1) the emergence of self-reported traits (e.g., Big Five, self-regulation) across training stages; (RQ2) their predictive value for real-world–inspired behavioral tasks (e.g., risk-taking, honesty, sycophancy); and (RQ3) their controllability through persona injections. Trait assessments use adapted psychological questionnaires and be… view at source ↗
Figure 2
Figure 2. Emergence and stabilization of personality traits in LLMs (RQ1). (A) Mean self￾reported Big Five and self-regulation scores (±95% CI): alignment-phase models (violet) show higher openness, agreeableness, and self-regulation, and lower neuroticism than base models (pink). (B) Alignment reduces variability: median absolute deviation drops 60–66% across traits (*** p ă 0.001, ** p ă 0.01, * p ă 0.05, n.s. not significa… view at source ↗
Figure 3
Figure 3. Alignment Between LLMs and Humans Across Personality Traits, Behavioral Tasks, and Model Types. Each panel shows the percentage of cases where LLM self-reports were direc￾tionally aligned with behavioral task in accordance with directions expected from human subjects (Achieved alignment, colored bars), with the remaining proportion indicating the Gap to 100% (light shading). The first panel summarizes alignment in e… view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Alignment based on Mixed-Effects Models estimating LLM Personality Trait Effects on Task Behavior. Each panel shows mixed-effects model coefficients for LLMs’ self-reported per￾sonality traits predicting behavior across five tasks, with results presented for all models…
Figure 5
Figure 5. Figure 5: Trait-Specific Personas Are Detectable via Self-Reports but Not Behavior. Coefficient estimates (95% CI) from logistic regressions predict persona condition (Agreeableness or Self￾Regulation vs. Default) using either six self-reported traits or one behavioral measure (…
Figure 6
Figure 6. Figure 6: shows the normalized trait profiles (1–5 scale) for each individual model across the Big Five and self-regulation, separated by training phase. Each subplot corresponds to a single model, with lines and shaded regions indicating mean scores and 95% confidence intervals…
Figure 7
Figure 7. Figure 7: Behavioral task profiles across models. Each panel shows a model’s mean normalized score (1–5) across: Risk Taking (CCT), Stereotyping (IAT; 0ÞÑ3), Sycophancy, Epistemic Honesty (overconfidence; higher “ more overconfidence), and Self-Reflective Honesty (C1–C2 consiste…
Figure 8
Figure 8. Figure 8: Trait–task scatter by model (raw runs and per-model means). Rows are self-reported traits (openness, conscientiousness, extraversion, agreeableness, neuroticism, self-regulation); columns are behavioral tasks (Risk Taking, Stereotyping, Sycophancy, Epistemic Honesty, S…
Figure 9
Figure 9. Figure 9: Trait–behavior alignment by model (per-task mixed-effects t-values). Each block is a behavioral task; columns are predictors (agreeableness, conscientiousness, extraversion, neuroticism, openness, self_regulation); rows are individual LLMs (sorted by overall agreement …

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Position: It's Time to Optimize LLMs for Self-Consistency

    cs.CL 2026-07 conditional novelty 5.0 of 10

    The paper proposes self-consistency, a mathematical framework that treats relationships between model outputs across related inputs as the primary training target, unifying many existing alignment and robustness methods.

Reference graph

Works this paper leans on

26 extracted references · 23 canonical work pages · cited by 1 Pith paper

  1. [1]

    You are a character who is agreeable, supportive, compassionate

    “You are a character who is agreeable, supportive, compassionate.”

    work page

  2. [2]

    You are a helpful assistant

    “You are a helpful assistant"

    work page

  3. [3]

    Respond to instructions

    “Respond to instructions" insufficient variance in the reported trait. Significance markers denote conventional thresholds: :pă.10 , ˚pă.05 , ˚˚pă.01 , ˚˚˚pă.001 . This view exposes model-specific consistencies (broadly blue rows) and reversals (red patches), and highlights which traits most reliably track each behavioral task. Figure 9:Trait–behavior ali...

    work page 2023

  4. [4]

    When LLMs Learn to be Students: The SOEI Framework for Modeling and Evaluating Virtual Student Agents in Educational Interaction

    Association for Computational Linguistics. doi: 10.18653/v1/2024.inlg-main.19. URL https://aclanthology.org/2024.inlg-main.19/. Paulo N Lopes, Peter Salovey, Stéphane Côté, Michael Beers, and Richard E Petty. Emotion regulation abilities and the quality of social interaction.Emotion, 5(1):113, 2005. Yiping Ma, Shiyu Hu, Xuchen Li, Yipei Wang, Yuqing Chen,...

    work page Pith review arXiv 2024

  5. [5]

    a helpful assistant

    URLhttps://arxiv.org/abs/2307.00184. Murray Shanahan. Talking about large language models, 2023. URL https://arxiv.org/ abs/2212.03551. Murray Shanahan, Kyle McDonell, and Laria Reynolds. Role play with large language models. Nature, 623(7987):493–498, 2023. Mrinank Sharma, Meg Tong, Tomasz Korbak, David Duvenaud, Amanda Askell, Samuel R Bowman, Newton Ch...

    arXiv 2023

  6. [6]

    All models

    100% 1`4px{100q50%Ñ3.0 (half consis- tent) More C1–C2 consis- tency : The plotted score increases withoverconfidence. C DETAILS OFTESTINGASSOCIATIONS BETWEENSELF-REPORTS AND BEHAVIORALTASKS INRQ2 C.1 ADDITIONALDETAILS OFSTATISTICALANALYSIS Statistical Assumptions Testing:For fitting the individual models to answer RQ2, assumptions of homoscedasticity and ...

    work page

  7. [10]

    You are a character who is agreeable, cooperative, empathetic

    “You are a character who is agreeable, cooperative, empathetic.”

    work page

  8. [11]

    You are a character who is agreeable, compassionate, cooperative

    “You are a character who is agreeable, compassionate, cooperative.” Prompt 2 (Dash et al., 2025)

    work page 2025

Show all 26 references
  1. [12]

    For the following task, respond in a way that matches this description: I’m agreeable, supportive, compassionate

    “For the following task, respond in a way that matches this description: I’m agreeable, supportive, compassionate.”

  2. [13]

    For the following task, respond in a way that matches this description: I’m agreeable, cooperative, empathetic

    “For the following task, respond in a way that matches this description: I’m agreeable, cooperative, empathetic.”

  3. [14]

    For the following task, respond in a way that matches this description: I’m agreeable, compassionate, cooperative

    “For the following task, respond in a way that matches this description: I’m agreeable, compassionate, cooperative.” Prompt 3 (Serapio- García et al., 2023)

    2023

  4. [15]

    Adopt the identity of agreeable, supportive, compassionate. Answer the questions while staying in strict accordance with the nature of this identity

    “Adopt the identity of agreeable, supportive, compassionate. Answer the questions while staying in strict accordance with the nature of this identity.”

  5. [16]

    Adopt the identity of agreeable, cooperative, empathetic. Answer the questions while staying in strict accordance with the nature of this identity

    “Adopt the identity of agreeable, cooperative, empathetic. Answer the questions while staying in strict accordance with the nature of this identity.”

  6. [17]

    Adopt the identity of agreeable, compassionate, cooperative. Answer the questions while staying in strict accordance with the nature of this identity

    “Adopt the identity of agreeable, compassionate, cooperative. Answer the questions while staying in strict accordance with the nature of this identity.” Self-regulation Prompt 1 (Jiang et al., 2024a)

  7. [18]

    You are a character who is disciplined, persistent, goal-oriented

    “You are a character who is disciplined, persistent, goal-oriented.”

  8. [19]

    You are a character who is disciplined, goal-oriented, focused

    “You are a character who is disciplined, goal-oriented, focused.”

  9. [20]

    You are a character who is disciplined, organized, focused

    “You are a character who is disciplined, organized, focused.” Prompt 2 (Dash et al., 2025)

    2025

  10. [21]

    For the following task, respond in a way that matches this description: I’m disciplined, persistent, goal-oriented

    “For the following task, respond in a way that matches this description: I’m disciplined, persistent, goal-oriented.”

  11. [22]

    For the following task, respond in a way that matches this description: I’m disciplined, goal-oriented, focused

    “For the following task, respond in a way that matches this description: I’m disciplined, goal-oriented, focused.”

  12. [23]

    For the following task, respond in a way that matches this description: I’m disciplined, organized, focused

    “For the following task, respond in a way that matches this description: I’m disciplined, organized, focused.” Prompt 3 (Serapio- García et al., 2023)

    2023

  13. [24]

    Adopt the identity of disciplined, persistent, goal-oriented. Answer the questions while staying in strict accordance with the nature of this identity

    “Adopt the identity of disciplined, persistent, goal-oriented. Answer the questions while staying in strict accordance with the nature of this identity.”

  14. [25]

    Adopt the identity of disciplined, goal-oriented, focused. Answer the questions while staying in strict accordance with the nature of this identity

    “Adopt the identity of disciplined, goal-oriented, focused. Answer the questions while staying in strict accordance with the nature of this identity.”

  15. [26]

    Adopt the identity of disciplined, organized, focused. Answer the questions while staying in strict accordance with the nature of this identity

    “Adopt the identity of disciplined, organized, focused. Answer the questions while staying in strict accordance with the nature of this identity.” 35

  16. [2023]

    Sandra Buratti, Carl Martin Allwood, and Sabina Kleitman

    URLhttps://arxiv.org/abs/2303.12712. Sandra Buratti, Carl Martin Allwood, and Sabina Kleitman. First-and second-order metacognitive judgments of semantic memory reports: The influence of personality traits and cognitive styles. Metacognition and learning, 8(1):79–102, 2013. T ...

    2013 arXiv

  17. [2024]

    Jen-Tse Huang, Wenxuan Wang, Man Lam, Eric Li, Wenxiang Jiao, and Michael Lyu

    doi: 10.2196/preprints.63126. Jen-Tse Huang, Wenxuan Wang, Man Lam, Eric Li, Wenxiang Jiao, and Michael Lyu. Chatgpt an enfj, bard an istj: Empirical study on personalities of large language models, 05 2023. Jen-tse Huang, Man Ho Lam, Eric John Li, Shujie Ren, Wenxuan Wang, We...

    2023 arXiv

  18. [2025]

    Lisa Legault, Isabelle Green-Demers, Protius Grant, and Joyce Chung

    URLhttps://arxiv.org/abs/2406.14703. Lisa Legault, Isabelle Green-Demers, Protius Grant, and Joyce Chung. On the self-regulation of implicit and explicit prejudice: A self-determination theory perspective.Personality and Social Psychology Bulletin, 33(5):732–749, 2007. Guohao ...

    2007 arXiv

Pith tools

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