{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,13]],"date-time":"2026-03-13T08:14:06Z","timestamp":1773389646853,"version":"3.50.1"},"reference-count":46,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,3,13]],"date-time":"2026-03-13T00:00:00Z","timestamp":1773360000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Psychol."],"abstract":"<jats:sec>\n                    <jats:title>Background<\/jats:title>\n                    <jats:p>Large Language Models (LLMs) have demonstrated exceptional capability in textual emotion detection. However, LLM evaluations often treat the emotion taxonomy\u2014the \u201ccognitive ruler\u201d defining the emotional space\u2014as a neutral background variable. The extent to which taxonomic complexity moderates LLM performance remains underexplored.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Method<\/jats:title>\n                    <jats:p>This study systematically evaluates the impact of emotion taxonomy on GPT-5\u2019s annotation behavior. We constructed a dataset of 2,848 Chinese Weibo posts. Five human annotators and GPT-5 (zero-shot) labeled the data across five distinct taxonomies, each with varying levels of granularity: SemEval (4 classes), Ekman (6 classes), Chinese SevenEmotions (7 classes), Plutchik (8 classes), and GoEmotions (27 classes). A rigorous experimental design, including randomized ordering and washout periods, was implemented to minimize sequence effects. By comparing the results of GPT-5 and manual annotation, the analysis is conducted across three dimensions: performance, consistency, and bias patterns.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>Results reveal a significant \u201cgranularity paradox\u201d: GPT-5\u2019s performance is strongly negatively correlated with taxonomic complexity, with performance collapsing in fine-grained settings (GoEmotions). Crucially, we identified systematic misalignment mechanisms: (1) Consistency decay: Human-AI agreement significantly deteriorates as semantic boundaries blur in complex taxonomies; (2) Hyper-sensitivity bias: GPT-5 exhibits a tendency to over-interpret neutral texts as emotional, with false-positive rates increasing with taxonomy size; and (3) Arousal shift: The model consistently misclassifies low-arousal negative emotions (e.g., sadness) as high-arousal prototypes (e.g., fear\/anger), reflecting a valence-based rather than nuance-based inference logic. Notably, the indigenous SevenEmotions did not yield superior cultural alignment compared to Western taxonomies.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusion<\/jats:title>\n                    <jats:p>Our findings suggest that emotion taxonomies function as a critical hyperparameter that shapes the cognitive boundaries of GPT. While GPT shows promise, its reliability is compromised by complex taxonomies. Researchers must balance granular detail against model robustness when deploying LLMs for psychological analysis.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/fpsyg.2026.1786724","type":"journal-article","created":{"date-parts":[[2026,3,13]],"date-time":"2026-03-13T05:26:23Z","timestamp":1773379583000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Granularity paradox: how emotion taxonomies shape GPT-5\u2019s affective cognition and human-AI alignment"],"prefix":"10.3389","volume":"17","author":[{"given":"Fa","family":"Zhang","sequence":"first","affiliation":[{"name":"School of Management, Guangdong University of Science and Technology","place":["Dongguan, China"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Jian","family":"Chen","sequence":"additional","affiliation":[{"name":"School of Business, Beijing Institute of Technology","place":["Zhuhai, China"]}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"1965","published-online":{"date-parts":[[2026,3,13]]},"reference":[{"key":"ref1","first-page":"1","article-title":"A systematic testbed for evaluating emotion classification in large language models","volume":"13","author":"Altun","year":"2025","journal-title":"J. Sci. Rep. B"},{"key":"ref2","doi-asserted-by":"publisher","first-page":"2204","DOI":"10.1109\/TAFFC.2024.3419593","article-title":"A wide evaluation of ChatGPT on affective computing tasks","volume":"15","author":"Amin","year":"2024","journal-title":"IEEE Trans. Affect. Comput."},{"key":"ref3","doi-asserted-by":"publisher","first-page":"211","DOI":"10.1109\/TAFFC.2021.3075638","article-title":"Detecting mental disorders in social media through emotional patterns\u2014the case of anorexia and depression","volume":"14","author":"Arag\u00f3n","year":"2023","journal-title":"IEEE Trans. Affect. Comput."},{"key":"ref4","doi-asserted-by":"publisher","first-page":"829","DOI":"10.1038\/nrn1201","article-title":"Working memory: looking back and looking forward","volume":"4","author":"Baddeley","year":"2003","journal-title":"Nat. Rev. Neurosci."},{"key":"ref5","doi-asserted-by":"publisher","first-page":"7924","DOI":"10.48550\/ARXIV.2403.17612","article-title":"\u201cYou are an expert annotator\u201d: automatic best-worst-scaling annotations for emotion intensity modeling","volume":"1","author":"Bagdon","year":"2024","journal-title":"NAACL"},{"key":"ref6","first-page":"3523","author":"Belay","year":"2025"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2501.02532","article-title":"Evaluating large language models against human annotators in latent content analysis: sentiment, political leaning, emotional intensity, and sarcasm","author":"Bojic","year":"2025","journal-title":"arXiv"},{"key":"ref8","first-page":"2104","author":"Bostan","year":"2018"},{"key":"ref9","doi-asserted-by":"publisher","first-page":"1","DOI":"10.33497\/2021.summer.1","article-title":"Pr\u00e9cis: emotions: the basics","volume":"3","author":"Brady","year":"2021","journal-title":"J. Philo. Emotion"},{"key":"ref10","doi-asserted-by":"publisher","first-page":"450","DOI":"10.1109\/TCSS.2019.2956957","article-title":"A survey of sentiment analysis from social media data","volume":"7","author":"Chakraborty","year":"2020","journal-title":"IEEE Trans. Comput. Soc. Syst."},{"key":"ref11","doi-asserted-by":"publisher","first-page":"968","DOI":"10.1057\/s41599-024-03397-4","article-title":"Emotional contagion on social media and the simulation of intervention strategies after a disaster event: a modeling study","volume":"11","author":"Chu","year":"2024","journal-title":"Humanit. Soc. Sci. Commun."},{"key":"ref12","doi-asserted-by":"publisher","first-page":"70489","DOI":"10.1109\/ACCESS.2024.3401203","article-title":"Evaluating the effect of emotion models on the generalizability of text emotion detection systems","volume":"12","author":"De Le\u00f3n Langur\u00e9","year":"2024","journal-title":"IEEE Access"},{"key":"ref13","first-page":"4040","author":"Demszky","year":"2020"},{"key":"ref14","doi-asserted-by":"publisher","first-page":"169","DOI":"10.1080\/02699939208411068","article-title":"An argument for basic emotions","volume":"6","author":"Ekman","year":"1992","journal-title":"Cogn. Emot."},{"key":"ref15","doi-asserted-by":"publisher","first-page":"e2205767119","DOI":"10.1073\/pnas.2205767119","article-title":"An analysis of emotions and the prominence of positivity in #BlackLivesMatter tweets","volume":"119","author":"Field","year":"2022","journal-title":"Proc. Natl. Acad. Sci."},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2412.15993","article-title":"Fearful falcons and angry llamas: emotion category annotations of arguments by humans and LLMs","author":"Greschner","year":"2025","journal-title":"arXiv"},{"key":"ref17","doi-asserted-by":"publisher","first-page":"66","DOI":"10.3390\/bs7040066","article-title":"On the importance of both dimensional and discrete models of emotion","volume":"7","author":"Harmon-Jones","year":"2017","journal-title":"Behav. Sci."},{"key":"ref18","first-page":"97053","volume-title":"Advances in Neural Information Processing Systems","author":"Huang","year":"2024"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2307.10169","article-title":"Challenges and applications of large language models","author":"Kaddour","year":"2023","journal-title":"arXiv"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.31219\/osf.io\/sdnbh_v2","article-title":"Does GPT-4 decode richer architecture of emotion? Mapping dimensional structure with 99 emotions","author":"Ke","year":"2025","journal-title":"Open Science Framework"},{"key":"ref21","doi-asserted-by":"publisher","first-page":"20416695221131832","DOI":"10.1177\/20416695221131832","article-title":"Origin of the ease of association of color names: comparison between humans and AI","volume":"13","author":"Komatsu","year":"2022","journal-title":"I-Perception"},{"key":"ref22","first-page":"365","volume-title":"Lecture Notes in Computer Science","author":"Koptyra","year":"2023"},{"key":"ref23","first-page":"189","volume-title":"Augmented Intelligence in Healthcare: A pragmatic and Integrated Analysis","author":"Kumar","year":"2022"},{"key":"ref24","doi-asserted-by":"publisher","first-page":"159","DOI":"10.2307\/2529310","article-title":"The measurement of observer agreement for categorical data","volume":"33","author":"Landis","year":"1977","journal-title":"Biometrics"},{"key":"ref25","doi-asserted-by":"publisher","first-page":"799","DOI":"10.1146\/annurev-psych-010213-115043","article-title":"Emotion and decision making","volume":"66","author":"Lerner","year":"2015","journal-title":"Annu. Rev. Psychol."},{"key":"ref26","doi-asserted-by":"publisher","first-page":"104419","DOI":"10.1016\/j.ijdrr.2024.104419","article-title":"The characteristics and functionalities of citizen-led disaster response through social media: a case study of the #HenanFloodsRelief on Sina Weibo","volume":"106","author":"Li","year":"2024","journal-title":"Int. J. Disaster Risk Reduct."},{"key":"ref27","first-page":"5487","author":"Liu","year":"2024"},{"key":"ref28","first-page":"1","author":"Mohammad","year":"2018"},{"key":"ref29","doi-asserted-by":"publisher","first-page":"94009","DOI":"10.1109\/ACCESS.2025.3574629","article-title":"Benchmarking open-source large language models for sentiment and emotion classification in Indonesian tweets","volume":"13","author":"Nasution","year":"2025","journal-title":"IEEE Access"},{"key":"ref30","doi-asserted-by":"publisher","first-page":"2650","DOI":"10.21437\/interspeech.2024-2282","article-title":"From text to emotion: unveiling the emotion annotation capabilities of LLMs","author":"Niu","year":"2024","journal-title":"Interspeech"},{"key":"ref31","volume-title":"The Cognitive Structure of Emotions","author":"Ortony","year":"1990"},{"key":"ref32","volume-title":"Emotions in Social Psychology: Key Readings","author":"Parrott","year":"2001"},{"key":"ref33","doi-asserted-by":"publisher","first-page":"344","DOI":"10.1511\/2001.4.344","article-title":"The nature of emotions: human emotions have deep evolutionary roots, a fact that may explain their complexity and provide tools for clinical practice","volume":"89","author":"Plutchik","year":"2001","journal-title":"Am. Sci."},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.2309.03564","article-title":"Supervised learning and large language model benchmarks on mental health datasets: cognitive distortions and suicidal risks in Chinese social media","author":"Qi","year":"2023","journal-title":"arXiv"},{"key":"ref35","doi-asserted-by":"publisher","first-page":"310","DOI":"10.1037\/0022-3514.66.2.310","article-title":"Evidence for universality and cultural variation of differential emotion response patterning","volume":"66","author":"Scherer","year":"1994","journal-title":"J. Pers. Soc. Psychol."},{"key":"ref36","volume-title":"The Book of Rites (Li ji)","author":"Sheng","year":"2013"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2601.03267","article-title":"OpenAI GPT-5 system card","author":"Singh","year":"2025","journal-title":"arXiv"},{"key":"ref38","doi-asserted-by":"publisher","first-page":"257","DOI":"10.1207\/s15516709cog1202_4","article-title":"Cognitive load during problem solving: effects on learning","volume":"12","author":"Sweller","year":"1988","journal-title":"Cogn. Sci."},{"key":"ref39","doi-asserted-by":"publisher","first-page":"4","DOI":"10.1007\/s13278-025-01428-9","article-title":"Large language models (LLM) in computational social science: prospects, current state, and challenges","volume":"15","author":"Thapa","year":"2025","journal-title":"Soc. Netw. Anal. Min."},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2509.14922","article-title":"A comparative evaluation of large language models for Persian sentiment analysis and emotion detection in social media texts","author":"Tohidi","year":"2025","journal-title":"arXiv"},{"key":"ref41","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1162\/coli_a_00461","article-title":"Dimensional modeling of emotions in text with appraisal theories: corpus creation, annotation reliability, and prediction","volume":"49","author":"Troiano","year":"2023","journal-title":"Comput. Ling."},{"key":"ref42","author":"Wang","year":"2020"},{"key":"ref43","doi-asserted-by":"publisher","first-page":"619","DOI":"10.1007\/s00357-019-9307-0","article-title":"Comparing the utility of different classification schemes for emotive language analysis","volume":"36","author":"Williams","year":"2019","journal-title":"J. Classif."},{"key":"ref44","doi-asserted-by":"publisher","first-page":"503","DOI":"10.1186\/s40359-024-02008-w","article-title":"Evaluation of emotion classification schemes in social media text: an annotation-based approach","volume":"12","author":"Zhang","year":"2024","journal-title":"BMC Psychol."},{"key":"ref45","doi-asserted-by":"crossref","first-page":"3881","DOI":"10.18653\/v1\/2024.findings-naacl.246","volume-title":"Findings of the Association for Computational Linguistics: NAACL 2024","author":"Zhang","year":"2024"},{"key":"ref46","doi-asserted-by":"publisher","first-page":"1066628","DOI":"10.3389\/fpsyg.2022.1066628","article-title":"China public emotion analysis under normalization of COVID-19 epidemic: using Sina Weibo","volume":"13","author":"Zhang","year":"2023","journal-title":"Front. Psychol."}],"container-title":["Frontiers in Psychology"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fpsyg.2026.1786724\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,13]],"date-time":"2026-03-13T05:26:27Z","timestamp":1773379587000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fpsyg.2026.1786724\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,3,13]]},"references-count":46,"alternative-id":["10.3389\/fpsyg.2026.1786724"],"URL":"https:\/\/doi.org\/10.3389\/fpsyg.2026.1786724","relation":{},"ISSN":["1664-1078"],"issn-type":[{"value":"1664-1078","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,3,13]]},"article-number":"1786724"}}