Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

PsychAdapter: Adapting LLM Transformers to Reflect Traits, Personality and Mental Health

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Injecting continuous personality scores into every transformer layer makes LLMs produce text that expert raters match to intended traits at 87 to 97 percent accuracy.

desk verdict Genuinely novel adapter for continuous trait control in LLMs, but the message-level teacher labels are the load-bearing risk; worth serious review. read the letter →

arxiv 2412.16882 v2 pith:4JFG4IY4 submitted 2024-12-22 cs.AI cs.CL

Huy Vu , Huy Anh Nguyen , Adithya V Ganesan , Swanie Juhng , Oscar N.E. Kjell , Joao Sedoc , Margaret L. Kern , Ryan L. Boyd
show 3 more authors
Lyle Ungar H. Andrew Schwartz Johannes C. Eichstaedt
This is my paper · ORCID
classification cs.AIcs.CL
keywords PsychAdapterpersonality-conditionedtextgenerationBigFivepersonalitymentalhealthlanguageparameter-efficientadaptationtransformermodelstext-basedpsychologicalassessmentcomputationalpsychology
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 claims that a lightweight add-on to an ordinary transformer language model can make it generate text reflecting specified personality, demographic, and mental-health profiles, with no prompting required. The key move is to accept a vector of continuous psychological scores and project it, via learned per-layer matrices, into the model's hidden states so the scores influence every layer during training and generation. Trained on social-media and blog text labeled by text-based assessment models, the resulting PsychAdapter produced language that expert raters matched to the intended Big Five levels 87.3% of the time and to intended depression and life-satisfaction levels 96.7% of the time. If the approach works as described, it offers prompt-free, continuous control of trait expression across transformer LLMs and opens a way to study psychological constructs through generated language.

What carries the argument

The load-bearing mechanism is a per-layer linear projection of the trait vector into the hidden-state key and value dimensions of an autoregressive transformer, with a separate learned matrix for each layer except the last. The trait vector is fed as the hidden state of a dummy begin-of-sentence token, so the psychological scores condition every layer during the standard next-word reconstruction objective. Training labels come from a teacher-student pipeline: participant-level text-based assessment models, built on topic or lexical features, are applied to each message to produce estimated Big Five, depression, life-satisfaction, or age scores, and the adapter learns to reconstruct the original message from that score vector. The added parameters are tiny relative to the base model, roughly 0.002% to 0.07% depending on the model, which is what allows PsychAdapters to be distributed and plugged into different base LLMs.

What would settle it

Train a PsychAdapter on message-level self-reported personality scores rather than teacher-model estimates and run the same expert-rater matching protocol: if accuracy falls to chance, the reported effect comes from the teacher distillation rather than from the per-layer conditioning architecture itself.

Watch

Extended reading notes

Core claim

The paper's central claim is that a transformer language model can be made to express a specified psychological profile by conditioning not on prompt words but on a continuous vector of trait scores, projected into the model's hidden states at every layer. PsychAdapter, the proposed modification, learns separate projection matrices for the key and value streams of each transformer layer so that a five-dimensional Big Five vector, or a depression, life-satisfaction, or age score, influences the representation at every layer. After training on messages labeled with scores from participant-level text-based assessment models, the adapter generates text that expert raters matched to the intended low, neutral, or high level with 87.3% average accuracy across Big Five traits and 96.7% for depression and life satisfaction; with age also conditioned, two expert raters reached 100% accuracy. The paper further reports that the same method transfers to GPT-2, Gemma, and Llama3 and produces text at five fine-grained trait levels, and that rotating extraversion and agreeableness inputs traces the interpersonal circumplex.

Load-bearing premise

The load-bearing premise is that the participant-level text-based assessment models, applied to individual messages to produce the trait labels the adapter trains on, give valid estimates of the intended psychological constructs at message level; if those estimates are wrong, the adapter learns the teacher models' linguistic stereotypes rather than genuine trait-language patterns.

Editorial extensions

If this is right

  • PsychAdapters make personality and mental-health traits a controllable input dimension of autoregressive LLMs, independent of prompt tokens, so trait expression no longer competes with the prompt for context-window space.
  • Chatbots and virtual agents could be instantiated with a specified Big Five profile, mental-health state, and age, enabling more human-like interaction and audience-matched content.
  • Clinical training tools could simulate conversation partners expressing depression or low life satisfaction without risk to patients.
  • Researchers can generate large, domain-focused text samples across the full continuous range of trait scores, supporting the study of how psychological constructs surface in language.
  • The same pipeline transfers to different base models, including GPT-2, Gemma, and Llama3, and across text domains such as tweets and blogs, suggesting the conditioning mechanism is model-agnostic.

Reading between the lines

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

  • One testable consequence is that the trait signal is bounded by the teacher estimates used to label messages; training the same architecture on message-level self-reported scores, where available, could produce measurably different linguistic patterns than training on participant-level model estimates projected onto messages.
  • Because the architecture only needs a continuous variable and a text-based teacher model for it, the same adapter design could in principle condition generation on other between-person dimensions such as political orientation, loneliness, or stress, not just the traits demonstrated in the paper.
  • If PsychAdapters are used to build 'digital participant' cohorts for psychological experiments, the generated cohort's trait-language distribution would inherit the teacher models' biases, so the validity of those cohorts as human proxies should be measured directly before relying on them.
  • A direct extension would be to interpolate continuously between trait levels and test whether human perceivers recover graded differences in single messages, rather than only the Low, Neutral, and High categories used in the reported evaluation.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 introduces PsychAdapter, a parameter-efficient modification of autoregressive transformer language models that injects continuous psychological trait vectors into every transformer layer via learned projection matrices and LoRA fine-tuning. Training uses message-level 'estimated' trait scores produced by participant-level language-based assessment models applied to blog and tweet corpora, with a text-reconstruction objective. The authors report that the resulting adapters generate text distinguishable by expert raters at 87.3% average accuracy for the Big Five, 96.7% for depression and life satisfaction, and that fine-grained levels, demographics, multi-trait combinations, and different base models (GPT-2, Gemma, Llama 3) can be controlled, with Claude 3.5 Sonnet used as an additional annotator.

Significance. If the central claim holds, PsychAdapter is a useful and lightweight contribution: it offers prompt-free, continuous control over trait expression in LLMs with under 0.1% added parameters, and the paper ships open code and data plus human expert evaluations. The architecture idea of per-layer trait projections is clean and clearly described. However, the validity of the entire pipeline rests on the unvalidated message-level teacher labels, and the headline accuracy figures are point estimates from a small sample without significance tests or confidence intervals. These are fixable concerns, so the contribution is potentially publishable, but the current evidence does not yet establish the strong psychological claim.

major comments (3)
  1. [Methods — 'Obtaining estimated psychological scores for messages'] The training signal for every PsychAdapter is the message-level estimate psi_m = W X_m, obtained by applying participant-level assessment models to individual messages. No validation is reported that these participant-level models, trained on aggregated text with LDA topic features, produce valid trait estimates at the message level for 30-word blog excerpts or 12-word tweets. If these labels capture only topical or lexical stereotypes, the adapter learns the teacher's stereotypes rather than genuine trait-language patterns, and the human-rater results would then measure stereotype matching rather than psychological construct expression. This is load-bearing for the paper's central claim: please add a validation study comparing message-level teacher estimates against human trait judgments (or against participant-level ground truth on held-out authors), and report per-construct correlations, agreement, and examples where teacher estimates diverge from human judgment.
  2. [Results — 'Evaluations with human raters and Claude'] The headline accuracies (87.3% and 96.7%) are point estimates based on 10 generation seeds, two raters, and a nonstandard scoring rule (one point only if all three levels are correctly matched; 1/3 point if exactly one is correct). No confidence intervals, significance tests, or per-dimension inter-rater reliabilities are reported, so the statement that PsychAdapter 'significantly outperformed' the 33.3% baseline is unsupported. Please report exact binomial or permutation tests against chance, CIs for each dimension and condition, and the raw trial-level rater classifications; also clarify whether raters were forced to assign one group to each level or could make other assignments.
  3. [Results — 'Generating text at fine-grained levels of personality' and Figures 4–5] The five-level fine-grained evaluation, which is the main evidence for continuous control, is conducted entirely with Claude 3.5 Sonnet as the annotator, with no human validation of the five-level distinctions. The earlier human–Claude agreement is established only at the coarse three-level task. Since the fine-grained claim is central to the advantage over discrete prompting, either add human ratings for the five-level setup or explicitly reframe the Claude-based results as a stereotype-matching measure rather than evidence of psychological validity.
minor comments (5)
  1. [Abstract] The abstract contains two typos: 'Lamma' for 'Llama' and 'PsychAdapter — es empirically derived trait-language patterns' which appears to be missing the word 'uses'.
  2. [Methods] The section heading 'Obtaining estimated psychological for messages' is missing the word 'scores'; please correct it.
  3. [Discussion — Limitations] The Limitations paragraph contains 'the propagation of bias' and 'a the propagation', which should be edited; more importantly, the paragraph does not mention the message-level teacher-label validity concern, which is the main methodological risk.
  4. [Figure 5 caption] The caption contains 'avaraged' for 'averaged' and would benefit from reporting the number of trials used for the reported accuracies.
  5. [Methods — 'Modifying transformer language models...'] The description of transformation matrices says 'each transformation matrix will have the shape of [latent_size, num_key_value_heads × head_dim]', but the text earlier refers to a 5 × 2048 matrix for Gemma-2B; please make the shape notation and the example consistent.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: external human ratings anchor the evaluation; teacher-label validity is a construct risk, not a circular loop.

full rationale

The derivation chain is: participant-level language-based assessment models, cited from prior peer-reviewed work, are applied at the message level to produce teacher labels psi_m = W X_m (Methods, 'Obtaining estimated psychological for messages'); PsychAdapter is trained to maximize p(text | psi); and generated text is evaluated by blinded Ph.D.-level psychologists (and by Claude, which was validated against those psychologists). The evaluation is an external benchmark, not the same fitted labeler, so the 87.3% / 96.7% accuracies are not forced by construction. The paper transparently labels the teacher outputs as 'estimated' scores and does not claim to validate the teacher models in this work. The main risk—that message-level teacher estimates may encode lexical stereotypes rather than the intended constructs—is a construct-validity and correctness concern, not a circularity: no equation in the paper reduces the claimed prediction to the fitted inputs by construction. The self-citations (Schwartz 2013, Park 2015, Schwartz 2014/2016) are load-bearing for the training labels but are externally grounded, peer-reviewed empirical models, not uniqueness theorems or unverified ansätze. The Limitations section acknowledges demographic bias in the training data but does not address teacher-label validity; that omission is a correctness risk, not evidence of circularity. Overall, the central evaluation is self-contained against external human judgment, so the circularity score is low.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central method depends on a chain of fitted models: the teacher assessment models define the labels, the projection matrices and LoRA weights are fit to those labels, and the evaluation is run on the outputs. The only externally grounded inputs are pretrained weights, the blog and tweet corpora, and human rater judgments of generated text.

free parameters (6)
  • Per-layer trait projection matrices W_trans_l for keys and values = 55,296 added parameters for Gemma-2B; 552,960 for GPT-2 Large; 393,216 for Llama-3-8B
    Learned during training to map the continuous trait vector into the transformer hidden states at each layer; this is the central mechanism of PsychAdapter.
  • LoRA low-rank matrices = r=8, alpha=32, dropout=0.1 on q/k/v/o/gate/up/down projections
    Used to fine-tune the base model while training the adapter; these hyperparameters are chosen by hand and are not reported as swept or optimized.
  • Teacher model weights for personality, depression, life satisfaction, and age = Not provided in the paper
    Pretrained regression models from prior work produce the estimated trait scores used as training labels; their weights are fitted to external data but are not shipped or independently validated at the message level here.
  • Trait intensity levels k = -3, -1.5, 0, 1.5, 3
    Chosen for the evaluation to represent Very Low, Low, Neutral, High, and Very High; these are interpretability choices, not fitted values.
  • Circumplex rotation angle alpha = 22.5 degrees
    Taken from DeYoung et al. to map warmth and dominance onto extraversion and agreeableness; not fitted in this paper.
  • Blog post truncation length = First 30 words
    Data preprocessing choice that shapes the training distribution and the style of generated text.
assumptions (5)
  • domain assumption A participant-level language-based assessment model can be applied to individual messages to produce valid psychological trait scores.
    The entire training set is labeled by applying participant-level regression weights to message-level word frequencies (Methods, 'Obtaining estimated psychological scores'), with no message-level validation reported.
  • domain assumption Big Five, depression, life satisfaction, and age are adequately represented by the specific text-based predictive models used.
    The paper relies on Schwartz 2013, Park 2015, prior depression and well-being models, and Sap 2014 as ground truth for the constructs; there is no clinical or self-report validation of the generated outputs.
  • domain assumption Trait expression in generated text can be measured by expert raters and by Claude 3.5 Sonnet as valid proxies.
    The evaluation protocol assumes blinded human raters and LLM annotations capture intended trait levels; Claude is used for large-scale evaluations after a limited human agreement check.
  • ad hoc to paper The text reconstruction objective with injected trait vectors is sufficient to learn trait-conditioned generation.
    There is no proof or ablation establishing that per-layer key/value projection is the correct or minimal conditioning mechanism; it is an architectural design choice.
  • domain assumption Injecting projected trait vectors at all layers except the last does not degrade the base model's language quality.
    The paper reports no fluency, coherence, or quality evaluation of the generated text, despite the modified hidden states across all layers.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PsychAdapter: Adapting LLM Transformers to Reflect Traits, Personality and Mental Health." pith.science (2026). https://pith.science/paper/4JFG4IY4

@misc{pith2026241216882,
  author       = {Pith},
  title        = {Pith review of: PsychAdapter: Adapting LLM Transformers to Reflect Traits, Personality and Mental Health},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4JFG4IY4}},
  note         = {Machine review of arXiv:2412.16882}
}
read the original abstract

Artificial intelligence-based language generators are now a part of most people's lives. However, by default, they tend to generate "average" language without reflecting the ways in which people differ. Here, we propose a lightweight modification to the standard language model transformer architecture - "PsychAdapter" - that uses empirically derived trait-language patterns to generate natural language for specified personality, demographic, and mental health characteristics (with or without prompting). We applied PsychAdapters to modify OpenAI's GPT-2, Google's Gemma, and Meta's Llama 3 and found generated text to reflect the desired traits. For example, expert raters evaluated PsychAdapter's generated text output and found it matched intended trait levels with 87.3% average accuracy for Big Five personalities, and 96.7% for depression and life satisfaction. PsychAdapter is a novel method to introduce psychological behavior patterns into language models at the foundation level, independent of prompting, by influencing every transformer layer. This approach can create chatbots with specific personality profiles, clinical training tools that mirror language associated with psychological conditionals, and machine translations that match an authors reading or education level without taking up LLM context windows. PsychAdapter also allows for the exploration psychological constructs through natural language expression, extending the natural language processing toolkit to study human psychology.

Figures

Figures reproduced from arXiv: 2412.16882 by the authors.

Figure 1
Figure 1. (A) Overview of PsychAdapter augmenting the widely-used auto-regressive transformers architecture2 to incorporate personality scores (associated with text) as additional inputs together with prior words and learn to reconstruct text. (B) After training, the base language model augmented with PsychAdapter can take different inputs reflecting any specific personality pattern and generates text accordingly. The output … view at source ↗
Figure 2
Figure 2. (A) The set-up for human expert evaluation. For each variable, PsychAdapter generated a group of text from 3 input levels: Low, Neutral, and High. The blinded expert evaluator attempted to match the output text to the input level. (B-D) Results for human evaluation tasks, measured by accuracy as the percentage of correct matches, compared against the random-chance baseline (33.3%). Generating text for mental health … view at source ↗
Figure 3
Figure 3. (A) Generated text controlling both mental health variables (e.g., depression, life satisfaction) and demographics variables (e.g., age). (B) Text generated around the interpersonal circumplex. Extraversion and agreeableness can be rotated to yield values for dominance and warmth32 and, thus, positions in the circumplex. Text for the desired circumplex positions was generated by using the corresponding values of ext… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Accuracy of correspondence between five intended levels of PsychGenerator output (modifying Gemma-2B) and annotations provided by Claude 3.5 Sonnet, for five traitlevels [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Accuracy between intended levels and annotated levels across LLMs. PsychAdapter was applied to (A) GPT-2 Large and (B) Llama3-8B, levels were annotated by Claude 3.5 Sonnet for levels: Very Low, Low, Neutral, High, and Very High. Results are based on 10 generated sampl…

Discussion (0). Continue with ORCID 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. Reward Valuation in Vision Language Models: Causal Mechanisms Underlying Anhedonia

    cs.LG 2026-07 conditional novelty 6.5 of 10

    Targeted perturbation of reward-anticipatory units in VLMs induces anhedonia-like effort avoidance and clinical-scale score drops without impairing baseline task competence.

Reference graph

Works this paper leans on

69 extracted references · 48 canonical work pages · cited by 1 Pith paper

  1. [1]

    Vaswani, A. et al. Attention is all you need. In Advances in Neural Information Processing Systems, vol. 30, 6000–6010 (2017)

    work page 2017

  2. [2]

    Radford, A. et al. Language models are unsupervised multitask learners. Technical Report, OpenAI (2019)

    work page 2019

  3. [3]

    Raffel, C. et al. Exploring the limits of transfer learning with a unified text-to-text transformer. J. Mach. Learn. Res. 21, 1–67 (2020)

    work page 2020

  4. [4]

    Lewis, M. et al. BART: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, 7871–7880, DOI: 10.18653/v1/2020.acl-main.703 (Online, 2020)

    doi:10.18653/v1/2020.acl-main.703

  5. [5]

    Zellers, R. et al. Defending against neural fake news. In Wallach, H. et al. (eds.) Advances in Neural Information Processing Systems, vol. 32, 9054–9065 (Curran Associates, Inc., 2019)

    work page 2019

  6. [6]

    i slept like a baby

    Giorgi, S. et al. " i slept like a baby": Using human traits to characterize deceptive chatgpt and human text. In IACT@SIGIR-2023: Special Interest Group on Information Retrieval (2023)

    work page 2023

  7. [7]

    Tak, A. N. & Gratch, J. Gpt-4 emulates average-human emotional cognition from a third-person perspective.arXiv preprint arXiv:2408.13718 (2024)

    arXiv 2024

  8. [8]

    the consistent lack of variance of psychological factors expressed by llms and spambots

    Varadarajan, V ., Giorgi, S. & Schwartz, H. A. "the consistent lack of variance of psychological factors expressed by llms and spambots". In under review (2025)

    work page 2025

Show all 69 references
  1. [9]

    W., Mehl, M

    Pennebaker, J. W., Mehl, M. R. & Niederhoffer, K. G. Psychological aspects of natural language use: Our words, our selves. Annu. Rev. Psychol. 54, 547–577 (2003)

    2003

  2. [10]

    Introducing the next generation of claude

    Anthropic. Introducing the next generation of claude. Anthropic Blog (2024)

    2024

  3. [11]

    Schwartz, H. A. et al. Personality, gender, and age in the language of social media: The open-vocabulary approach. PloS one 8 (2013). 16/19

    2013

  4. [12]

    Schwartz, H. A. et al. Dlatk: Differential language analysis toolkit (2017)

    2017

  5. [13]

    Schwartz, H. A. et al. Towards assessing changes in degree of depression through Facebook. In Proceedings of the Workshop on Computational Linguistics and Clinical Psychology: From Linguistic Signal to Clinical Reality, 118–125, DOI: 10.3115/v1/W14-3214 (Association for Comput...

    doi

  6. [14]

    Eichstaedt, J. C. et al. Facebook language predicts depression in medical records. Proc. Natl. Acad. Sci. 115, 11203–11208 (2018)

    2018

  7. [15]

    Kern, M. L. et al. The online social self: An open vocabulary approach to personality. Assessment 21, 158–169 (2014)

    2014

  8. [16]

    Park, G. et al. Automatic personality assessment through social media language. J. Pers. Soc. Psychol. 108, 934–952 (2015)

    2015

  9. [17]

    Kern, M. L. et al. Gaining insights from social media language: Methodologies and challenges. Psychol. Methods (2016)

    2016

  10. [18]

    Schwartz, H. et al. Characterizing geographic variation in well-being using tweets. Proc. Int. AAAI Conf. on Web Soc. Media 7, 583–591, DOI: 10.1609/icwsm.v7i1.14442 (2021)

    2021 doi

  11. [19]

    & Walker, M

    Mairesse, F. & Walker, M. PERSONAGE: Personality generation for dialogue. In Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics , 496–503 (Association for Computational Linguistics, Prague, Czech Republic, 2007)

    2007

  12. [20]

    & Walker, M

    Mairesse, F. & Walker, M. A. Controlling user perceptions of linguistic style: Trainable generation of personality traits. Comput. Linguist. 37, 455–488, DOI: 10.1162/COLI_a_00063 (2011)

    2011 doi

  13. [21]

    & Konopnicki, D

    Herzig, J., Shmueli-Scheuer, M., Sandbank, T. & Konopnicki, D. Neural response generation for customer service based on personality traits. In Proceedings of the 10th International Conference on Natural Language Generation, 252–256, DOI: 10.18653/v1/W17-3541 (Association for C...

    doi

  14. [22]

    & Mao, X

    Zheng, Y ., Zhang, R., Huang, M. & Mao, X. A pre-training based personalized dialogue generation model with persona- sparse data. Proc. AAAI Conf. on Artif. Intell. 34, 9693–9700, DOI: 10.1609/aaai.v34i05.6518 (2020)

    2020 doi

  15. [23]

    & Zhu, X

    Qian, Q., Huang, M., Zhao, H., Xu, J. & Zhu, X. Assigning personality/profile to a chatting machine for coherent conversa- tion generation. In Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence, IJCAI-18, 4279–4285, DOI: 10.24963/ijcai....

    doi

  16. [24]

    Li, J. et al. A persona-based neural conversation model. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , 994–1003, DOI: 10.18653/v1/P16-1094 (Association for Computational Linguistics, Berlin, Germany, 2016)

    doi

  17. [25]

    Jiang, H. et al. Personallm: Investigating the ability of large language models to express personality traits. arXiv preprint arXiv:2305.02547 (2023)

    2023 arXiv

  18. [26]

    Safdari, M. et al. Personality traits in large language models. arXiv preprint arXiv:2307.00184 (2023)

    2023 arXiv

  19. [27]

    Jiang, G. et al. Evaluating and inducing personality in pre-trained language models. InProceedings of the 37th International Conference on Neural Information Processing Systems, NIPS ’23 (Curran Associates Inc., Red Hook, NY , USA, 2024)

    2024

  20. [28]

    & Srivastava, S

    Caron, G. & Srivastava, S. Identifying and manipulating the personality traits of language models. arXiv preprint arXiv:2212.10276 (2022)

    2022 arXiv

  21. [29]

    Liu, H. et al. Few-shot parameter-efficient fine-tuning is better and cheaper than in-context learning. In Proceedings of the 36th International Conference on Neural Information Processing Systems, NIPS ’22 (Curran Associates Inc., Red Hook, NY , USA, 2024)

    2024

  22. [30]

    Eichstaedt, J. C. et al. Closed-and open-vocabulary approaches to text analysis: A review, quantitative comparison, and recommendations. Psychol. Methods 26, 398 (2021)

    2021

  23. [31]

    Sap, M. et al. Developing age and gender predictive lexica over social media. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), 1146–1151, DOI: 10.3115/v1/D14-1121 (2014)

    2014 doi

  24. [32]

    McCrae, R. R. & Costa, P. T. The structure of interpersonal traits: Wiggins’s circumplex and the five-factor model. J. Pers. Soc. Psychol. 56, 586–595, DOI: 10.1037/0022-3514.56.4.586 (1989)

    1989 doi

  25. [33]

    Wiggins, J. S. A psychological taxonomy of trait-descriptive terms: The interpersonal domain. J. Pers. Soc. Psychol. 37, 395–412 (1979)

    1979

  26. [34]

    Wiggins, J. S. Circumplex models of interpersonal behavior in clinical psychology. Handb. research methods clinical psychology 183–221 (1982). 17/19

    1982

  27. [35]

    G., Weisberg, Y

    DeYoung, C. G., Weisberg, Y . J., Quilty, L. C. & Peterson, J. B. Unifying the aspects of the big five, the interpersonal circumplex, and trait affiliation. Journals Pers. 81, 465–475 (2013)

    2013

  28. [36]

    Schwartz, H. A. et al. Women are warmer but no less assertive than men: Gender and language on facebook. PLOS ONE 11, e0155885, DOI: 10.1371/journal.pone.0155885 (2016)

    2016 doi

  29. [37]

    Touvron, H. et al. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971 (2023)

    2023 arXiv

  30. [38]

    & Lee, K

    Nass, C. & Lee, K. Does computer-generated speech manifest personality? an experimental test of similarity-attraction. Proc. SIGCHI Conf. on Hum. Factors Comput. Syst.329–336 (2000)

    2000

  31. [39]

    & Yan, C

    Lee, K., Peng, W., Jin, S.-A. & Yan, C. Can robots manifest personality?: An empirical test of personality recognition, social responses, and social presence in human–robot interaction. J. Commun. 56, 754–772 (2006)

    2006

  32. [40]

    & Dautenhahn, K

    Fong, T., Nourbakhsh, I. & Dautenhahn, K. A survey of socially interactive robots robotics and autonomous systems. Robotics Auton. Syst. 42, 143–166 (2003)

    2003

  33. [41]

    & Castellano, G

    Paetzel-Prüsmann, M., Perugia, G. & Castellano, G. The influence of robot personality on the development of uncanny feelings. Comput. Hum. Behav. 120, 106756, DOI: https://doi.org/10.1016/j.chb.2021.106756 (2021)

    2021

  34. [42]

    & Johnson, L

    Shumanov, M. & Johnson, L. Making conversations with chatbots more personalized. Comput. Hum. Behav. 117, 106627, DOI: https://doi.org/10.1016/j.chb.2020.106627 (2021)

    2021

  35. [43]

    Zhang, K., Lu, G., Zhang, G., Lei, Z. & Wu, L. Personalized headline generation with enhanced user interest perception. In Pimenidis, E., Angelov, P., Jayne, C., Papaleonidas, A. & Aydin, M. (eds.) Artificial Neural Networks and Machine Learning – ICANN 2022, 797–809 (Springer...

    2022

  36. [44]

    Ao, X. et al. PENS: A dataset and generic framework for personalized news headline generation. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long ...

    doi

  37. [45]

    & Lee, J

    Kim, D. & Lee, J. Designing an algorithm-driven text generation system for personalized and interactive news reading. Int. J. Human-Computer Interact. 35, 109–122 (2019)

    2019

  38. [46]

    & Willer, R

    Hewitt, L., Ashokkumar, A., Ghezae, I. & Willer, R. Predicting results of social science experiments using large language models (2024). Working Paper

    2024

  39. [47]

    & Mervielde, I

    Van Hiel, A., Kossowska, M. & Mervielde, I. The relationship between openness to experience and political ideology. Pers. Individ. Differ. 28, 741–751 (2000)

    2000

  40. [48]

    Sutskever, I., Vinyals, O. & Le, Q. V . Sequence to sequence learning with neural networks. In Proceedings of the 27th International Conference on Neural Information Processing Systems - Volume 2, NIPS’14, 3104–3112 (MIT Press, Cambridge, MA, USA, 2014)

    2014

  41. [49]

    & Fernández, R

    Xing, Y . & Fernández, R. Automatic evaluation of neural personality-based chatbots. In Krahmer, E., Gatt, A. & Goudbeek, M. (eds.) Proceedings of the 11th International Conference on Natural Language Generation, 189–194, DOI: 10.18653/v1/W18-6524 (Association for Computationa...

    doi

  42. [50]

    Oraby, S. et al. Controlling personality-based stylistic variation with neural natural language generators. In Komatani, K. et al. (eds.) Proceedings of the 19th Annual SIGdial Meeting on Discourse and Dialogue, 180–190, DOI: 10.18653/v1/ W18-5019 (Association for Computationa...

    doi

  43. [51]

    & Wang, D

    Zhou, Z., Shen, Y ., Chen, X. & Wang, D. Gerp: A personality-based emotional response generation model.Appl. Sci. 13, DOI: 10.3390/app13085109 (2023)

    2023 doi

  44. [52]

    & Graves, A

    Graves, A. & Graves, A. Long short-term memory. Supervised sequence labelling with recurrent neural networks 37–45 (2012)

    2012

  45. [53]

    Social media fact sheet

    Pew Research Center. Social media fact sheet. Tech. Rep., Pew Research Center (2023). Accessed on November 13, 2024

    2023

  46. [54]

    & Pennebaker, J

    Schler, J., Koppel, M., Argamon, S. & Pennebaker, J. Effects of age and gender on blogging. 199–205 (2006)

    2006

  47. [55]

    Giorgi, S. et al. The remarkable benefit of user-level aggregation for lexical-based population-level predictions. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing(2018)

    2018

  48. [56]

    Schwartz, H. A. et al. Predicting individual well-being through the language of social media (2016)

    2016

  49. [57]

    Team, G. et al. Gemma: Open models based on gemini research and technology. arXiv preprint arXiv:2403.08295 (2024). 18/19

    2024 arXiv

  50. [58]

    Adelani, D. I. et al. Generating sentiment-preserving fake online reviews using neural language models and their human- and machine-based detection. Adv. Inf. Netw. Appl. 1341–1354 (2019)

    2019

  51. [59]

    & Pal, C

    Pilault, J., Li, R., Subramanian, S. & Pal, C. On extractive and abstractive neural document summarization with transformer language models. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), 9308–9319, DOI: 10.18653/v1/2020.emnl...

    2020 doi

  52. [60]

    & Shaikh, S

    Santhanam, S. & Shaikh, S. Emotional neural language generation grounded in situational contexts. In Proceedings of the 4th Workshop on Computational Creativity in Language Generation, 22–27 (Tokyo, Japan, 2019)

    2019

  53. [61]

    & Delangue, C

    Wolf, T., Sanh, V ., Chaumond, J. & Delangue, C. Transfertransfo: A transfer learning approach for neural network based conversational agents (2019). 1901.08149

    2019 arXiv

  54. [62]

    DIALOGPT : Large-scale generative pre-training for conversational response generation

    Zhang, Y .et al. DIALOGPT : Large-scale generative pre-training for conversational response generation. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics: System Demonstrations , 270–278, DOI: 10.18653/v1/2020.acl-demos.30 (Online, 2020)

    doi

  55. [63]

    S., McCann, B., Varshney, L., Xiong, C

    Keskar, N. S., McCann, B., Varshney, L., Xiong, C. & Socher, R. CTRL - A Conditional Transformer Language Model for Controllable Generation. arXiv preprint arXiv:1909.05858 (2019)

    2019 arXiv

  56. [64]

    & Johnson, M

    Wang, Y ., Wood, I., Wan, S., Dras, M. & Johnson, M. Mention flags (MF): Constraining transformer-based text generators. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language P...

    doi

  57. [65]

    & Seddah, D

    Jawahar, G., Sagot, B. & Seddah, D. What does BERT learn about the structure of language? In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 3651–3657, DOI: 10.18653/v1/P19-1356 (Association for Computational Linguistics, Florence, Italy, 2019)

    doi

  58. [66]

    Hu, E. J. et al. LoRA: Low-rank adaptation of large language models. In International Conference on Learning Representations (2022)

    2022

  59. [67]

    Matero, M. et al. Using daily language to understand drinking: Multi-level longitudinal differential language analysis. In Yates, A. et al. (eds.) Proceedings of the 9th Workshop on Computational Linguistics and Clinical Psychology (CLPsych 2024), 133–144 (Association for Comp...

    2024

  60. [68]

    Yaden, D. B. et al. Characterizing empathy and compassion using computational linguistic analysis. Emotion 24, 106–115, DOI: 10.1037/emo0001205 (2023)

    2023 doi

  61. [69]

    Pennebaker, J. W. & King, L. A. Linguistic styles: Language use as an individual difference. J. Pers. Soc. Psychol. 77, 1296–1312 (1999). Acknowledgements We thank Todd N. Karhu for helpful comments on ethical implications. This work was supported by NIH grants R01AA028032, R0...

    1999

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.