Generative artificial intelligence (AI) models will increasingly replace humans in producing output for a variety of important tasks. While much prior work has mostly focused on the improvement in the average performance of generative AI models relative to humans’ performance, much less attention has been paid to the significant reduction of variance in output produced by generative AI models. In this article, we demonstrate that generative AI models are inherently prone to the phenomenon of “regression toward the mean”, whereby variance in output tends to shrink relative to that in real-world distributions. We discuss potential social implications of this phenomenon across three levels—societal, group, and individual—and two dimensions—material and non-material. Finally, we discuss interventions to mitigate negative effects, considering the roles of both service providers and users. Overall, this article aims to raise awareness of the importance of output variance in generative AI and to foster collaborative efforts to meet the challenges posed by the reduction of variance in output generated by AI models.
AbidAFarooqiMZouJ (2021Persistent anti-Muslim bias in large language models, proceedings of the 2021 AAAI/ACM conference on AI, Ethics, and Society, virtual event, USA, pp.298–306. New York, NY, USA: Association for Computing Machinery. Available at:https://doi.org/10.1145/3461702.3462624.
Baidoo-AnuDAnsahLO (2023) Education in the era of generative artificial intelligence (AI): Understanding the potential benefits of ChatGPT in promoting teaching and learning. Journal of AI7(1): 52–62.
4.
BianchiFKalluriPDurmusE, et al. (2023) Easily accessible text-to-image generation amplifies demographic stereotypes at large scale. In: proceedings of the 2023 ACM conference on fairness, accountability, and transparency, pp.1493–1504. Chicago, IL, USA: Association for Computing Machinery. Available at:https://doi.org/10.1145/3593013.35940.
5.
BurtonJWLopez-LopezEHechtlingerS, et al. (2024) How large language models can reshape collective intelligence. Nature Human Behaviour8(9): 1643–1655.
6.
ChoiASAkterSSSinghJP, et al. (2024) The LLM effect: Are humans truly using LLMs, or are they being influenced by them instead? In: Al-Onaizan Y, Bansal M and Chen Y-N (eds) proceedings of the 2024 conference on empirical methods in natural language processing, Miami, Florida, USA, pp.22032–22054. Association for Computational Linguistics. Available at:https://aclanthology.org/2024.emnlp-main.1230/.
7.
ChowdheryANarangSDevlinJ, et al. (2023) Palm: Scaling language modeling with pathways. Journal of Machine Learning Research24(240): 1–113.
8.
ClementCBBierbaumMKeeffeO, et al. (2019) On the use of arXiv as a dataset. arXiv preprint, arXiv:1905.00075.
9.
ColeJRColeS (1972) The Ortega hypothesis: Citation analysis suggests that only a few scientists contribute to scientific progress. Science178(4059): 368–375.
10.
CroitoruF-AHondruVIonescuRT, et al. (2023) Diffusion models in vision: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence45(9): 10850–10869.
11.
DarwinCMurrayJSonsWCA, et al. (1859) On the Origin of Species by Means of Natural Selection, or, The Preservation of Favoured Races in the Struggle for Life. London: John Murray. Pdf. retrieved from the Library of Congress, available at: https://www.loc.gov/item/06017473/.
12.
DickinsonDLOaxacaRL (2009) Statistical discrimination in labor markets: An experimental analysis. Southern Economic Journal76(1): 16–31.
13.
EtxanizJAzkuneGSoroaA, et al. (2024) BertaQA: How much do language models know about local culture? In: Globerson A, Mackey L, Belgrave D, et al. (eds) advances in neural information processing systems 37, Vancouver, BC, Canada, 10–15 December 2024, pp.34077–34097. Vancouver, Canada: Neural Information Processing Systems Foundation, Inc.
14.
FanALewisMDauphinY (2018) Hierarchical neural story generation. In: Gurevych I and Miya Y (eds) proceedings of the 56th annual meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Melbourne, Australia, July 2018, pp.889–898. Association for Computational Linguistics.
15.
FerraraE (2023) Should ChatGPT be biased? Challenges and risks of bias in large language models. arXiv preprint, arXiv:2304.03738.
16.
FiskeSTCuddyAJCGlickP, et al. (2018) A model of (often mixed) stereotype content: competence and warmth respectively follow from perceived status and competition. Journal of Personality and Social Psychology 82(6): 878–902.
17.
GallegosIORossiRABarrowJ, et al. (2024) Bias and fairness in large language models: A survey. Computational Linguistics1: 1–79.
GaoJWangD (2024) Quantifying the use and potential benefits of artificial intelligence in scientific research. Nature Human Behaviour8(12): 2281–2292.
20.
GoodfellowIPouget-AbadieJMirzaM, et al. (2020) Generative adversarial networks. Communications of the ACM63(11): 139–144.
21.
GuoZJinRLiuC, et al. (2023) Evaluating large language models: A comprehensive survey. arXiv preprint, arXiv:2310.19736.
22.
HackerPMittelstadtBZuiderveen BorgesiusF, et al. (2024) Generative discrimination: What happens when generative AI exhibits bias, and what can be done about it. arXiv preprint, arXiv:2407.10329.
23.
HoltzmanABuysJDuL, et al. (2020) The curious case of neural text degeneration. In: 8th international conference on learning representations, ICLR 2020, Addis Ababa, Ethiopia, 26–30 April 2020. OpenReview.net.
24.
HongLPageSE (2004) Groups of diverse problem solvers can outperform groups of high-ability problem solvers. Proceedings of the National Academy of Sciences101(46): 16385–16389.
25.
HongLPageSE (2012) Some microfoundations of collective wisdom. In: LandemoreHElsterJ (eds) Collective Wisdom. Cambridge: Cambridge University Press, pp.56–71.
26.
HuangZYuPAllanJ (2023) Improving cross-lingual information retrieval on low-resource languages via optimal transport distillation. In: proceedings of the sixteenth ACM international conference on web search and data mining, Singapore, pp.1048–1056. New York, NY, USA: Association for Computing Machinery.
27.
JinYChandraMVermaG, et al. (2024) Better to ask in English: cross-lingual evaluation of large language models for healthcare queries. Nature Machine Intelligence 6(4): 383–392.
28.
KanaharaS (2006) A review of the definitions of stereotype and a proposal for a progressional model. Individual Differences Research4(5): 306–321.
29.
KandpalNDengHRobertsA, et al. (2023) Large language models struggle to learn long-tail knowledge. In: Krause A, Brunskill E, Cho K, et al. (eds) International conference on machine learning, pp.15696–15707. Honolulu, Hawaii, USA: MLR.org.
30.
KirkHRVidgenBRöttgerP, et al. (2024) The benefits, risks and bounds of personalizing the alignment of large language models to individuals. Nature Machine Intelligence6(4): 383–392.
31.
KlangELevy-MendelovichS (2023) Evaluation of OpenAI’s large language model as a new tool for writing papers in the field of thrombosis and hemostasis. Journal of Thrombosis and Haemostasis21(4): 1055–1058.
32.
KotekHDockumRSunD (2023) Gender bias and stereotypes in large language models. In: Bernstein M, Savage S and Bozzon A (eds) proceedings of the ACM collective intelligence conference, Delft, Netherlands, pp.12–24. New York, NY, USA: Association for Computing Machinery. Available at:https://doi.org/10.1145/3582269.3615599.
33.
KungTHCheathamMMedenillaA, et al. (2023) Performance of ChatGPT on USMLE: Potential for AI-assisted medical education using large language models. PLoS Digital Health2(2): e0000198.
34.
LeeBCChungJ (2024) An empirical investigation of the impact of ChatGPT on creativity. Nature Human Behaviour8(10): 1906–1914.
35.
LiuJLiuCZhouP, et al. (2023) LLMRec: Benchmarking large language models on recommendation task. arXiv preprint, arXiv:2305.15114.
36.
LiuYYangKQiZ, et al. (2024) Bias and volatility: A statistical framework for evaluating large language model’s stereotypes and the associated generation inconsistency. In: Globerson A, Mackey L, Belgrave D, et al. (eds) advances in neural information processing systems 37, Vancouver, BC, Canada, pp.110131–110155. 10–15 December 2024. Neural Information Processing Systems Foundation, Inc.
37.
LoCK (2023) What is the impact of ChatGPT on education? A rapid review of the literature. Education Sciences13(4): 10.
38.
MayrE (1982) The Growth of Biological Thought: Diversity, Evolution, and Inheritance. Cambridge, MA: Harvard University Press.
PageS (2008) The Difference: How the Power of Diversity Creates Better Groups, Firms, Schools, and Societies-New Edition. Princeton, NJ: Princeton University Press.
42.
PetersonAJ (2025) AI and the problem of knowledge collapse. AI & Society40(5): 3249–3269.
RothsteinDSCarrDCookseyE (2019) Cohort profile: The national longitudinal survey of youth 1979 (NLSY79). International Journal of Epidemiology48(1): 22–22e.
45.
SawyerRKHenriksenD (2024) Explaining Creativity: The Science of Human Innovation. 3rd ed.Oxford: Oxford University Press.
46.
SiCYangDHashimotoT (2024) Can LLMs generate novel research ideas? A large-scale human study with 100+ NLP researchers. arXiv preprint, arXiv:2409.03770.
47.
SpencerSJLogelCDaviesPG (2016) Stereotype threat. Annual Review of Psychology67: 415–437.
48.
SteeleCMAronsonJ (1995) Stereotype threat and the intellectual test performance of African Americans. Journal of Personality and Social Psychology69(5): 797–811.
49.
TouvronHLavrilTIzacardG, et al. (2023a) Llama: Open and efficient foundation language models. arXiv preprint, arXiv:2302.13971.
50.
TouvronHMartinLStoneK, et al. (2023b) Llama 2: Open foundation and fine-tuned chat models. arXiv preprint, arXiv:2307.09288.
51.
TverskyAKahnemanD (1974) Judgment under uncertainty: Heuristics and biases: Biases in judgments reveal some heuristics of thinking under uncertainty. Science185(4157): 1124–1131.
52.
WanYPuGSunJ, et al. (2023) “Kelly is a warm person, Joseph is a role model": Gender biases in LLM-generated reference letters. arXiv preprint, arXiv:2310.09219.
53.
XieY (2013) Population heterogeneity and causal inference. Proceedings of the National Academy of Sciences110(16): 6262–6268.
54.
XieYQiTYiJ, et al. (2024) Measuring human contribution in AI-assisted content generation. arXiv preprint, arXiv: 2403.16875.
55.
ZhaoWXZhouKLiJ, et al. (2023) A survey of large language models. arXiv preprint, arXiv:2303.18223.
56.
ZhouMAbhishekVDerdengerT, et al. (2024) Bias in generative AI. arXiv preprint, arXiv:2407.10853.
