Pith. sign in

REVIEW 3 major objections 5 minor 40 references

Injecting extracted MBTI and OCEAN personality traits into prompts plus contrastive retrieval of similar scenarios measurably improves how well large language models infer emotions and their causes.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-14 13:13 UTC pith:IX2IM5S4

load-bearing objection Clean engineering that gets a few real points on EmoBench by stuffing MBTI+OCEAN into prompts and retrieval; the unvalidated LLM trait labels are the softest link, but the gains and ablations are honest. the 3 major comments →

arxiv 2607.10245 v1 pith:IX2IM5S4 submitted 2026-07-11 cs.CL cs.IR

PTEI: Integrating Personality Traits to Enhance Emotional Intelligence in Large Language Models

Amir Reza Jafari , Praboda Rajapaksha , Reza Farahbakhsh , Noel Crespi This is my paper
classification cs.CL cs.IR
keywords Emotional IntelligenceEmotion Detection and AnalysisLanguage ModelingPersonality TraitsMBTIOCEANLarge Language ModelsContrastive Learning
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Large language models still lag humans on complex emotional understanding because they treat every scenario the same, ignoring that different people feel different things in the same situation. The paper argues that the missing piece is personality: enduring traits that shape how humans interpret and react to events. PTEI first extracts MBTI type and OCEAN trait levels directly from a short scenario, then feeds those traits into carefully built prompts together with a few retrieved examples that match both the emotion and the personality profile. Contrastive learning is used to train an embedding space so that retrieval surfaces genuinely similar cases. On the EmoBench Emotional Understanding benchmark the method lifts accuracy for several models, with the largest gains on GPT-4o when personality-aware prompts are combined with chain-of-thought reasoning, producing roughly a four-point overall improvement. A sympathetic reader cares because the result shows a concrete, psychologically grounded route toward AI systems that reason about feelings the way people actually do, rather than as generic averages.

Core claim

Conditioning large language models on MBTI and OCEAN personality traits extracted from an emotional scenario, and retrieving few-shot examples that are aligned in both emotion and personality via a contrastively trained embedding space, improves emotion-and-cause prediction accuracy on the EmoBench Emotional Understanding task, with the strongest gains appearing when the same personality context is combined with chain-of-thought prompting on GPT-scale models.

What carries the argument

PTEI: a pipeline that (1) extracts MBTI type and five OCEAN levels from the scenario text, (2) builds a synthetic memory bank of 500 personality-annotated emotional scenarios, (3) fine-tunes a sentence encoder with a personality-aware contrastive (NT-Xent) objective so that similar emotion-plus-personality pairs cluster, and (4) retrieves top-k aligned examples to condition both base and chain-of-thought prompts for joint emotion-cause prediction.

Load-bearing premise

The MBTI and OCEAN labels that a small language model extracts from short synthetic or benchmark scenarios are accurate and stable enough proxies for the real psychological differences that drive the gold emotion and cause answers.

What would settle it

Replace the automatically extracted personality labels with human-expert or questionnaire-derived MBTI and OCEAN profiles for the same EmoBench scenarios and re-run the exact PTEI-Base and PTEI-CoT evaluations; if the accuracy gains disappear, the central claim fails.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes PTEI, a framework that extracts MBTI types and OCEAN trait levels from emotional scenarios via few-shot LLM prompting, then injects them into personality-aware few-shot and CoT prompts. A synthetic 500-scenario memory bank (GPT-4 generated, dual-LLM + human filtered, κ=0.92) is embedded with a personality-aware contrastive objective (NT-Xent) so that top-k retrieval supplies emotionally and personally aligned exemplars. On the 200-item EmoBench EU task, PTEI-Base and especially PTEI-CoT improve overall accuracy for several LLMs, with the largest lift for GPT-4o (PTEI-CoT 63.62 % vs. standard CoT 58.88 %). Ablations isolate contributions of traits versus retrieval; robustness checks use majority voting and option permutation.

Significance. If the reported gains hold under broader evaluation, the work supplies a practical, modular recipe for conditioning LLM emotional reasoning on individual-difference signals and demonstrates that both trait strings and contrastively retrieved exemplars contribute. Strengths include the dual-framework (MBTI+OCEAN) design, the carefully filtered synthetic bank, explicit ablations (Table 2, Figure 3), and majority-vote/permutation robustness. The absolute lifts remain modest and model-scale dependent, yet the approach is immediately usable by practitioners and opens a clear line of inquiry into psychologically grounded prompting for EI benchmarks.

major comments (3)
  1. [Section 3.5 / Limitations] Section 3.5 (Personality Detection Module) and the Limitations section: MBTI and OCEAN labels are produced solely by GPT-4o-mini few-shot prompts applied to the identical short scenario text later used for EU inference. No human agreement, inter-annotator reliability, correlation with any external personality instrument, or even consistency checks across re-prompts are reported for either the 200 EmoBench items or the 500 synthetic scenarios. Consequently the central claim that gains arise from “psychologically grounded” personality integration rests on an unvalidated proxy; the TraitsOnly ablation (Table 2) shows only that a structured string helps, not that the string encodes latent personality factors that drive the gold labels. External validation or at least a controlled human-agreement study on a subset is load-bearing for the interpretive narrative.
  2. [Section 4.2 / Table 1] Section 4.2 and Table 1: Evaluation rests on only 200 English multiple-choice EU items. No statistical significance tests, confidence intervals, or power analysis accompany the reported point improvements (e.g., +4.74 points for GPT-4o PTEI-CoT). With this N, category-level fluctuations (especially PT and PBE) and the mixed or negative effects observed for smaller models (Qwen-7B PTEI-CoT) cannot be reliably distinguished from noise. Expanding the test suite or reporting bootstrap intervals is required before the “enhances EU capabilities of various LLMs” claim can be considered robust.
  3. [Sections 3.3–3.6] Sections 3.3–3.4 (Memory Bank) and 3.6 (Contrastive Learning): Both the memory-bank scenarios and the personality labels are generated by LLMs (GPT-4 / GPT-4o-mini). Although the authors carefully avoid lexical overlap with EmoBench and apply quality filters, the possibility remains that retrieval surfaces surface-form regularities already known to the same model family rather than genuine emotion–personality alignments. A control that replaces the contrastive library with random or emotion-only retrieval (beyond the existing RAG-only ablation) would isolate whether the personality-aware embedding space is doing the claimed work.
minor comments (5)
  1. [Figure 1] Figure 1 caption and body text contain typos (“Underestanding”, “Categoty”).
  2. [Abstract / §5.1] Abstract and §5.1 claim an “additional 4 percent increase”; Table 1 shows +4.74 points for GPT-4o PTEI-CoT over CoT—please align the rounded figure with the table.
  3. [Eq. (9) / §3.6] Equation (9) defines S_ps with α=0.5 by default; sensitivity of retrieval quality and downstream accuracy to α is never reported, yet α is listed among free parameters.
  4. [Appendix A] Appendix figures (emotion distribution, MBTI/OCEAN correlations) are referenced but their quantitative takeaways are not integrated into the main analysis; a short paragraph linking them to the observed category-level gains would strengthen interpretability.
  5. [§4 / Appendix C] Prompt templates in Appendix C are useful; however, the exact number of retrieved shots (k) used at inference is never stated in the main text or tables.

Circularity Check

0 steps flagged

No significant circularity: reported EU accuracy gains rest on independent EmoBench gold labels and are not equivalent by construction to the LLM-inferred personality inputs or synthetic memory bank.

full rationale

The paper is an empirical prompting-and-retrieval framework, not a first-principles derivation. Personality traits (MBTI/OCEAN) are extracted from scenario text via GPT-4o-mini few-shot prompts (Section 3.5) and injected as context; the memory bank of 500 synthetic scenarios is generated by GPT-4 and annotated with the same module, then used only for contrastive embedding training and few-shot retrieval (Sections 3.3–3.6). Final evaluation, however, measures accuracy against the fixed, human-validated gold emotion/cause labels of the 200 EmoBench EU items (Table 1, Section 4.2). Ablations (Table 2, Figure 3) isolate the contribution of traits versus retrieval, and no equation, uniqueness claim, or self-citation chain reduces the reported lifts (e.g., GPT-4o PTEI-CoT 63.62 % vs. CoT 58.88 %) to a tautology of the inputs. The Limitations section itself flags the proxy nature of the inferred traits, confirming that validity is an open empirical question rather than a definitional circularity. No self-definitional loops, fitted-parameter-as-prediction, load-bearing self-citations, imported uniqueness theorems, or ansatz smuggling appear in the derivation chain.

Axiom & Free-Parameter Ledger

5 free parameters · 3 axioms · 2 invented entities

The central accuracy claims rest on a handful of hand-chosen similarity thresholds and weighting coefficients, on the assumption that LLM-inferred MBTI/OCEAN labels are useful proxies, and on the synthetic memory bank whose quality is itself judged largely by LLMs. No free parameters are fitted to the final EmoBench test labels, but several design choices are fixed without sensitivity analysis beyond the reported ablations.

free parameters (5)
  • personality similarity weight α = 0.5
    Balances MBTI vs OCEAN in the composite similarity used for both contrastive pair construction and retrieval; default 0.5 chosen by authors.
  • positive similarity threshold θ_s = 0.7
    Scenarios with S_ps ≥ θ_s and matching emotion become positive pairs; set to 0.7.
  • negative dissimilarity threshold θ_d = 0.3
    Scenarios with S_ps < θ_d become negative pairs even if emotion matches; set to 0.3.
  • NT-Xent temperature τ = 0.07
    Temperature in the contrastive loss; fixed at the common default 0.07.
  • quality acceptance thresholds τ_q = τ_h = 3.5
    Minimum average LLM-judge and human Likert scores for retaining a synthetic scenario; both set to 3.5.
axioms (3)
  • domain assumption MBTI types and OCEAN trait levels inferred from short textual scenarios by GPT-4o-mini are sufficiently faithful proxies of the psychological factors that modulate emotion and cause labels.
    Invoked throughout Section 3.5 and the personality-aware prompts; Limitations section acknowledges the approximation.
  • ad hoc to paper Synthetic EU scenarios generated by GPT-4 under the authors’ category template are distributionally close enough to EmoBench items that retrieval from the synthetic bank improves real-test performance without leakage.
    Section 3.3–3.4; authors claim no textual overlap, but the claim is not independently verified.
  • domain assumption The four EmoBench EU categories (CE, PBE, PT, EC) adequately sample the space of complex emotional understanding.
    Evaluation is performed exclusively on these 200 items (Section 4.2).
invented entities (2)
  • PTEI framework (personality detection + contrastive embedding library + personality-aware Base/CoT prompts) no independent evidence
    purpose: To condition LLM emotion/cause inference on extracted personality traits and retrieved analogues.
    The entire pipeline is introduced by the paper; no prior system combines both trait frameworks with this retrieval objective for EU.
  • 500-scenario personality-enriched synthetic memory bank no independent evidence
    purpose: Source of positive/negative pairs for contrastive learning and of few-shot exemplars at inference.
    Constructed ad hoc for this work; quality rests on LLM judges and a 10 % human sample.

pith-pipeline@v1.1.0-grok45 · 22656 in / 3124 out tokens · 29510 ms · 2026-07-14T13:13:51.859724+00:00 · methodology

0 comments
read the original abstract

Despite advances in Emotional Intelligence (EI), Large Language Models (LLMs) still significantly underperform humans in complex emotional reasoning. This gap originates partly from the limited incorporation of individual differences, particularly personality traits, which are fundamental to human emotional inference. To address this, we propose PTEI, a novel framework for integrating Personality Traits into Emotional Intelligence tasks using LLMs. In PTEI, MBTI and OCEAN personality traits are first extracted directly from the given emotional scenarios and then utilized as contextual knowledge within personality-aware prompts, guiding LLMs to accurately infer emotions and their underlying causes. To ensure optimal contextual grounding, we employ Contrastive Learning to construct an optimized retrieval system that surfaces emotionally and personally aligned scenarios, enhancing reasoning quality. Extensive experiments on established EI benchmarks show that PTEI enhances the Emotional Understanding (EU) capabilities of various LLMs, with the strongest improvement observed in GPT models. Combining PTEI with Chain-of-Thought (CoT) reasoning yields an additional 4 percent increase in accuracy. These findings underscore PTEI's contribution toward advancing AI systems with more sophisticated social and psychological grounding.

Figures

Figures reproduced from arXiv: 2607.10245 by Amir Reza Jafari, Noel Crespi, Praboda Rajapaksha, Reza Farahbakhsh.

Figure 1
Figure 1. Figure 1: Illustration of PTEI’s impact on emotional inference [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: Average accuracy under different personality input [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Case study comparison for GPT-4o, showing how personality-aware prompting improves emotional prediction [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Flattened Emotion Category Distribution (Normalized) [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: MBTI vs Emotion Correlation • Complex Emotions (CB): This task category evaluates a model’s ability to reason about nuanced and layered emo￾tional experiences. It includes scenarios involving (1) emo￾tional transitions in response to evolving events, (2) mixtures of emotions with potentially conflicting valence (e.g., joy and disappointment), and (3) unexpected emotional outcomes that challenge commonsense… view at source ↗
Figure 7
Figure 7. Figure 7: OCEAN Traits vs Emotion Correlation culturally influenced norms (e.g., attitudes toward punctual￾ity), differing levels of sentimental value assigned to objects or events, or reactions driven by personal traits and experi￾ences, such as phobias or personality styles. Accurate pre￾diction in this category requires sensitivity to how internal beliefs and past experiences modulate emotional appraisal. • Persp… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

40 extracted references · 10 canonical work pages

  1. [1]

    Mohammad Hossein Amirhosseini and Hassan Kazemian. 2020. Machine learn- ing approach to personality type prediction based on the myers–briggs type indicator®.Multimodal Technologies and Interaction4, 1 (2020), 9. doi:10.3390/ MTI4010009

    2020

  2. [2]

    Jinze Bai, Shuai Bai, Yunfei Chu, Zeyu Cui, Kai Dang, Xiaodong Deng, Yang Fan, Wenbin Ge, Yu Han, Fei Huang, and et al. 2023. Qwen Technical Report.arXiv preprint arXiv:2309.16609(2023). doi:10.48550/arXiv.2309.16609

    Pith/arXiv arXiv doi:10.48550/arxiv.2309.16609 2023

  3. [3]

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. 2020. A Simple Framework for Contrastive Learning of Visual Representations. In Proceedings of the 37th International Conference on Machine Learning. PMLR, 1597–1607. doi:10.5555/3524938.3525087

    doi:10.5555/3524938.3525087 2020

  4. [4]

    Zohar Elyoseph, Dorit Hadar-Shoval, Kfir Asraf, and Maya Lvovsky. 2023. Chat- GPT Outperforms Humans in Emotional Awareness Evaluations.Frontiers in Psychology14 (2023), 1199058. doi:10.3389/fpsyg.2023.1199058

    work page doi:10.3389/fpsyg.2023.1199058 2023

  5. [5]

    Yumeng Fu, Junjie Wu, Zhongjie Wang, Meishan Zhang, Lili Shan, Yulin Wu, and Bingquan Liu. 2025. LaERC-S: Improving LLM-based Emotion Recognition in Conversation with Speaker Characteristics. InProceedings of the 31st International Conference on Computational Linguistics, Owen Rambow, Leo Wanner, Marianna Apidianaki, Hend Al-Khalifa, Barbara Di Eugenio, a...

    2025

  6. [6]

    Daniel Goleman. 1996. Emotional intelligence. Why it can matter more than IQ. Learning24, 6 (1996), 49–50

    1996

  7. [7]

    Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Ab- hishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, and et al. 2024. The Llama 3 Herd of Models.arXiv preprint arXiv:2407.21783(2024). doi:10.48550/arXiv.2407.21783 PTEI: Integrating Personality Traits to Enhance Emotional Intelligence in Large Lan...

    Pith/arXiv arXiv doi:10.48550/arxiv.2407.21783 2024

  8. [8]

    James D Hess and Arnold C Bacigalupo. 2011. Enhancing decisions and decision- making processes through the application of emotional intelligence skills.Man- agement decision49, 5 (2011), 710–721. doi:10.1108/00251741111130805

    work page doi:10.1108/00251741111130805 2011

  9. [9]

    Izard, Deborah Z

    Carroll E. Izard, Deborah Z. Libero, Priscilla Putnam, and O. Maurice Haynes

  10. [10]

    Journal of Personality and Social Psychology64, 5 (May 1993), 847–860

    Stability of Emotion Experiences and Their Relations to Traits of Personality. Journal of Personality and Social Psychology64, 5 (May 1993), 847–860. doi:10.1 037//0022-3514.64.5.847

    1993

  11. [11]

    Jeff Johnson, Matthijs Douze, and Hervé Jégou. 2021. Billion-Scale Similarity Search with GPUs. 535-547 pages. doi:10.1109/TBDATA.2019.2921572

    doi:10.1109/tbdata.2019.2921572 2021

  12. [12]

    Yang Li, Amirmohammad Kazemeini, Yash Mehta, and Erik Cambria. 2022. Mul- titask learning for emotion and personality traits detection.Neurocomputing493 (2022), 340–350. doi:10.1016/J.NEUCOM.2022.04.049

    work page doi:10.1016/j.neucom.2022.04.049 2022

  13. [13]

    Zheng Li, Sujian Li, Dawei Zhu, Qilong Ma, and Weimin Xiong. 2025. EERPD: Leveraging Emotion and Emotion Regulation for Improving Personality De- tection. InProceedings of the 31st International Conference on Computational Linguistics, Owen Rambow, Leo Wanner, Marianna Apidianaki, Hend Al-Khalifa, Barbara Di Eugenio, and Steven Schockaert (Eds.). Associat...

    2025

  14. [14]

    Mahmudul Haque, Zakaria Masud Jiyad, M

    Abdullah Al Maruf, Fahima Khanam, Md. Mahmudul Haque, Zakaria Masud Jiyad, M. F. Mridha, and Zeyar Aung. 2024. Challenges and Opportunities of Text-Based Emotion Detection: A Survey.IEEE Access12 (2024), 18416–18450. doi:10.1109/ACCESS.2024.3356357

    work page doi:10.1109/access.2024.3356357 2024

  15. [15]

    McCrae and Paul T

    Robert R. McCrae and Paul T. Jr. Costa. 1997. Personality Trait Structure as a Human Universal.American Psychologist52, 5 (May 1997), 509–516. doi:10.1037/ /0003-066x.52.5.509

    1997

  16. [16]

    Robert R McCrae and Oliver P John. 1992. An introduction to the five-factor model and its applications.Journal of personality60, 2 (1992), 175–215. doi:10.1 111/j.1467-6494.1992.tb00970.x

    arXiv 1992

  17. [17]

    Yash Mehta, Navonil Majumder, Alexander Gelbukh, and Erik Cambria. 2020. Recent trends in deep learning based personality detection.Artificial Intelligence Review53, 4 (2020), 2313–2339. doi:10.1007/S10462-019-09770-Z

    work page doi:10.1007/s10462-019-09770-z 2020

  18. [18]

    1987.Introduction to type: A description of the theory and applications of the Myers-Briggs Type Indicator

    Isabel Briggs Myers. 1987.Introduction to type: A description of the theory and applications of the Myers-Briggs Type Indicator. Consulting Psychologists Press

    1987

  19. [19]

    OpenAI. 2023. GPT-4 Technical Report.arXiv preprint arXiv:2303.08774(2023). doi:10.48550/arXiv.2303.08774

    Pith/arXiv arXiv doi:10.48550/arxiv.2303.08774 2023

  20. [20]

    Samuel J Paech. 2023. Eq-bench: An emotional intelligence benchmark for large language models.arXiv preprint arXiv:2312.06281(2023). doi:10.48550/ARXIV.2 312.06281

    Pith/arXiv arXiv doi:10.48550/arxiv.2 2023

  21. [21]

    Nils Reimers and Iryna Gurevych. 2019. Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks. InProceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Confer- ence on Natural Language Processing (EMNLP-IJCNLP), Kentaro Inui, Jing Jiang, Vincent Ng, and Xiaojun Wan (Eds.). Association...

    doi:10.18653/v1/d19-1410 2019

  22. [22]

    Michael D Robinson and Gerald L Clore. 2002. Belief and feeling: Evidence for an accessibility model of emotional self-report.Psychological Bulletin128, 6 (2002), 934–960. doi:10.1037/0033-2909.128.6.934

    work page doi:10.1037/0033-2909.128.6.934 2002

  23. [23]

    Sahand Sabour, Siyang Liu, Zheyuan Zhang, June Liu, Jinfeng Zhou, Alvionna Sunaryo, Tatia Lee, Rada Mihalcea, and Minlie Huang. 2024. EmoBench: Eval- uating the Emotional Intelligence of Large Language Models. InProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Lun-Wei Ku, Andre Martins, and V...

    doi:10.18653/v1/2024.acl-long.326 2024

  24. [24]

    Mustafa Safdari, Greg Serapio-García, Clément Crepy, Stephen Fitz, Peter Romero, Luning Sun, Marwa Abdulhai, Aleksandra Faust, and Maja J. Mataric. 2023. Personality Traits in Large Language Models.CoRRabs/2307.00184 (2023). arXiv:2307.00184 doi:10.48550/ARXIV.2307.00184

    Pith/arXiv arXiv doi:10.48550/arxiv.2307.00184 2023

  25. [25]

    Peter Salovey and John D. Mayer. 1990. Emotional Intelligence.Imagination, Cognition and Personality9, 3 (1990), 185–211. doi:10.2190/DUGG-P24E-52WK- 6CDG

    work page doi:10.2190/dugg-p24e-52wk- 1990

  26. [26]

    Maarten Sap, Ronan Le Bras, Daniel Fried, and Yejin Choi. 2022. Neural Theory-of-Mind? On the Limits of Social Intelligence in Large LMs. InPro- ceedings of the 2022 Conference on Empirical Methods in Natural Language Pro- cessing, Yoav Goldberg, Zornitsa Kozareva, and Yue Zhang (Eds.). Association for Computational Linguistics, Abu Dhabi, United Arab Emi...

    doi:10.18653/v1/2022.emnlp-main.248 2022

  27. [27]

    Schuller

    Dagmar Schuller and Björn W. Schuller. 2018. The Age of Artificial Emotional Intelligence.Computer51, 9 (2018), 38–46. doi:10.1109/MC.2018.3620963

    doi:10.1109/mc.2018.3620963 2018

  28. [28]

    Aleksandra Sorokovikova, Natalia Fedorova, Sharwin Rezagholi, and Ivan P Yamshchikov. 2024. Llms simulate big five personality traits: Further evidence. arXiv preprint arXiv:2402.01765(2024). doi:10.48550/ARXIV.2402.01765

    Pith/arXiv arXiv doi:10.48550/arxiv.2402.01765 2024

  29. [29]

    Sanja Stajner and Seren Yenikent. 2020. A Survey of Automatic Personality Detection from Texts. InProceedings of the 28th International Conference on Computational Linguistics, COLING 2020, Barcelona, Spain (Online), December 8 13,2020. International Committee on Computational Linguistics, 6284–6295. doi:10.18653/V1/2020.COLING-MAIN.553

    work page doi:10.18653/v1/2020.coling-main.553 2020

  30. [30]

    Chenwei Wan, Matthieu Labeau, and Chloé Clavel. 2025. EmoDynamiX: Emo- tional Support Dialogue Strategy Prediction by Modelling MiXed Emotions and Discourse Dynamics. InProceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), Luis Chiruz...

    work page doi:10.18653/v1/2025.naacl-long.81 2025

  31. [31]

    Xuena Wang, Xueting Li, Zi Yin, Yue Wu, and Jia Liu. 2023. Emotional Intelligence of Large Language Models.Journal of Pacific Rim Psychology17 (2023). doi:10.1 177/18344909231213958

    2023

  32. [32]

    Yan Wang, Bo Wang, Yachao Zhao, Dongming Zhao, Xiaojia Jin, Jijun Zhang, Ruifang He, and Yuexian Hou. 2024. Emotion Recognition in Conversation via Dynamic Personality. InProceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), Nicoletta Calzolari, Min-Yen Kan, Veronique Hos...

    2024

  33. [33]

    https://aclanthology.org/2024.lrec-main.507/

    2024

  34. [34]

    Jialiang Wu, Yi Shen, Ziheng Zhang, and Longjun Cai. 2024. Enhancing Large Language Model with Decomposed Reasoning for Emotion Cause Pair Extraction. arXiv preprint arXiv:2401.17716(2024). doi:10.48550/arXiv.2401.17716

    Pith/arXiv arXiv doi:10.48550/arxiv.2401.17716 2024

  35. [35]

    Tao Yang, Tianyuan Shi, Fanqi Wan, Xiaojun Quan, Qifan Wang, Bingzhe Wu, and Jiaxiang Wu. 2023. PsyCoT: Psychological Questionnaire as Powerful Chain- of-Thought for Personality Detection. InFindings of the Association for Com- putational Linguistics: EMNLP 2023, Houda Bouamor, Juan Pino, and Kalika Bali (Eds.). Association for Computational Linguistics, ...

    work page doi:10.18653/v1/2023.findings-emnlp.216 2023

  36. [36]

    Haein Yeo, Taehyeong Noh, Seungwan Jin, and Kyungsik Han. 2025. PADO: Personality-induced multi-Agents for Detecting OCEAN in human-generated texts. InProceedings of the 31st International Conference on Computational Lin- guistics, Owen Rambow, Leo Wanner, Marianna Apidianaki, Hend Al-Khalifa, Barbara Di Eugenio, and Steven Schockaert (Eds.). Association ...

    2025

  37. [37]

    Perspective Taking: False Belief, Faux Pas, Strange Story

  38. [38]

    Emotional Cues: Vocal Cues, Visual Cues

  39. [39]

    Complex Emotions: Emotion Transition, Mixture of Emotions, Unexpected Outcome

  40. [40]

    Goal" (selected from above) –

    Personal Beliefs and Experiences: Cultural Values, Sentimental Value, Persona Main Prompt Body: This situation should contain: – "Goal" (selected from above) – "Category" (with category description) – "Scenario" (short, coherent situation) – "Subject" (the person the scenario is about) – 5–6 "Choices" for Emotion (single or combined, from predefined list)...

    2017