Pith. sign in

REVIEW 4 major objections 5 minor 2 cited by

Generative AI and Creativity: A Systematic Literature Review and Meta-Analysis

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read GenAI matches humans on creativity, aids teams, cuts diversity

desk verdict First meta-analysis on GenAI and creativity: transparent and useful, but the headline null effect rests on a pseudo-replication problem and the publication-bias correction reverses it; still deserves peer review. read the letter →

arxiv 2505.17241 v1 pith:EZMAHNPG submitted 2025-05-22 cs.HC cs.AI

Niklas Holzner , Sebastian Maier , Stefan Feuerriegel This is my paper
classification cs.HCcs.AI
keywords generativeAIcreativitymeta-analysisHedges'ghuman-AIcollaborationideadiversitydivergentthinkingsystematicreview
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper asks whether generative AI can produce creative ideas and whether it helps or hurts human creativity, then answers with a meta-analysis of 28 experimental studies (8,214 participants, 127 effect sizes). It finds that GenAI alone is, on average, neither more nor less creative than unaided humans (Hedges' $g = -0.05$, not significant), while humans who collaborate with GenAI are modestly more creative than those who do not ($g = 0.27$). It also finds that such collaboration substantially reduces the diversity of the ideas people generate ($g = -0.86$). These three results together frame GenAI as an augmentation tool that can support individual ideation while posing a homogenization risk, rather than as a replacement for human creativity. A careful reader should care because this is a pooled, quantitative picture of a scattered literature, and it separates model-, task-, and population-specific effects from a general claim.

What carries the argument

The carrying machinery is a random-effects meta-analysis of standardized effect sizes using Hedges' $g$. Effect sizes are computed from each eligible experiment either directly or by converting $t$-values, $F$-values, means, standard deviations, and $\beta$ coefficients with standard formulas, then pooled with a random-effects estimator designed for heterogeneous study sets. Heterogeneity is quantified with $Q$, $I^2$, and $\tau^2$, and moderator structure is probed with subgroup analyses, meta-regressions, leave-one-out sensitivity checks, regression tests for funnel asymmetry, and trim-and-fill bias adjustments. The design organizes the 127 effect sizes into three comparison families—human versus GenAI (RQ1), human plus GenAI versus human (RQ2a), and idea diversity in collaboration (RQ2b)—which is what lets a single study base produce three distinct answers.

What would settle it

Recompute the three pooled effects separately for each creativity-measure family: expert ratings, self-reports, semantic-distance scores, and rule-based scores. If the subgroup estimates differ enough to flip the sign or significance of any headline $g$ (for example, if the $g = 0.27$ collaboration gain disappears under expert ratings), then the different measures are not measuring one construct and the pooled numbers are not a single creativity effect.

Watch

Extended reading notes

Core claim

The paper's central claim is that the current empirical literature supports a three-part description of generative AI and creativity. First, in direct comparisons, GenAI-generated ideas are rated about as creative as human ideas (pooled Hedges' $g = -0.05$; 95% CI $[-0.26, 0.16]$), so average parity, not superhuman creativity, is the norm. Second, human-GenAI collaboration raises creative performance relative to unaided humans ($g = 0.27$; 95% CI $[0.02, 0.53]$), a small but robust effect that survives leave-one-out deletion and bias adjustments. Third, the same collaboration lowers the diversity of ideas ($g = -0.86$; 95% CI $[-1.33, -0.40]$), an effect the authors interpret as a homogenization risk. The authors also show that these averages hide real moderators: GPT-4 outperforms humans when generating alone, laypeople gain most from collaboration, and creative-writing tasks favor humans while divergent-thinking batteries favor GenAI. These results are the basis for the paper's conclusion that GenAI should be treated as an augmentative support, not a creative replacement.

Load-bearing premise

The analysis assumes that the different creativity measurements in the 28 studies—expert ratings, self-assessments, novelty scales, semantic distance, cosine similarity, and rule-based flexibility scores—are close enough to the same underlying construct that averaging their effect sizes into one pooled $g$ is meaningful.

Editorial extensions

If this is right

  • The pooled parity estimate ($g = -0.05$, CI $[-0.26, 0.16]$) means claims that GenAI is broadly more creative than humans are not supported by current evidence; any advantage is model-specific, such as GPT-4's $g = 0.50$.
  • The significant collaboration gain ($g = 0.27$) supports deploying GenAI as an ideation aid; the effect survives leave-one-out deletion and trim-and-fill adjustment.
  • The large diversity drop ($g = -0.86$, CI $[-1.33, -0.40]$) implies that using GenAI in creative work can homogenize outputs, a cost that should be weighed alongside the performance gain.
  • Heterogeneity results imply that creativity is not a single skill: GenAI does better on divergent-thinking batteries such as alternate uses and consequences tasks but worse on creative writing, and laypeople benefit more than academics.
  • The diversity estimate rests on six observations from four studies; its direction is robust to removing any single study, but its magnitude is not precise.

Reading between the lines

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

  • A joint interpretation the paper stops short of: the same GenAI that raises average creativity in collaboration could be doing so by anchoring people to a narrower idea space, so the performance gain and the diversity loss may be two sides of one mechanism rather than independent effects.
  • A testable extension: measure whether the collaboration gain shrinks as participants' domain expertise rises, ideally holding the GenAI model fixed; the paper's moderator pattern (laypeople gain, academics do not) predicts exactly that gradient.
  • A caution about generalizing: the diversity estimate is built on six observations from four studies, all text-based; extending the same meta-analytic comparison to image, music, and organizational ideation would show whether the homogenization effect is a general property of GenAI or a text-task property.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper reports a systematic literature review and meta-analysis (PRISMA 2020) of 28 studies examining whether GenAI matches human creative performance (RQ1), whether humans collaborating with GenAI outperform humans working alone on creative tasks (RQ2a), and whether such collaboration reduces idea diversity (RQ2b). Using random-effects meta-analysis with Hedges' g, the authors report no significant difference between GenAI and human creativity (g = -0.05), a significant positive collaboration effect (g = 0.27), and a significant negative diversity effect (g = -0.86). They also report subgroup analyses by model, participant background, and task type, along with robustness checks including leave-one-out, Egger's test, trim-and-fill, and influence diagnostics. Code and data are provided in a public repository.

Significance. If the three headline estimates withstand reanalysis, the paper would provide a useful quantitative synthesis of a fast-moving literature, with practical implications for human-AI collaboration. The authors deserve credit for making code and data publicly available, for following a structured PRISMA workflow, and for reporting multiple heterogeneity analyses. However, the current statistical implementation has two load-bearing weaknesses: a unit-of-analysis problem from treating 52 observations of one study as independent in RQ1, and a publication-bias analysis that reverses the RQ1 null result yet is described as raising no substantial concerns. These issues must be resolved before the central claims can be accepted.

major comments (4)
  1. [Section 3.2, Table 2] In the RQ1 analysis, HAI15 (Sun et al.) contributes 52 of the 100 observations, and the statement in Section 3.2 that 'each study contributes less than 1.1% weight' refers to per-observation weights. Because these 52 observations share one protocol, one participant pool, and one modeling pipeline, treating them as independent units inflates precision and can dominate the pooled g = -0.048. The leave-one-out analysis removes single observations and therefore cannot detect dependence on the entire HAI15 study. Please re-estimate RQ1 with study-level aggregation or a three-level/multivariate meta-analysis with cluster-robust variance estimation, and report both observation-level and study-level estimates.
  2. [Section 3.2.2, Section 2.3] Egger's test in Section 3.2.2 is significant (z = -3.363, p = 0.001) and the trim-and-fill procedure imputes 24 studies, shifting the pooled estimate to g = 0.364 with 95% CI [0.131, 0.596], thereby reversing the direction of the RQ1 effect. This contradicts the statement in Section 2.3 that the bias analyses 'revealed no substantial concerns regarding bias.' The paper must reconcile this contradiction, present the bias-corrected estimate prominently, and temper the conclusion that GenAI and humans show 'parity' in creative performance.
  3. [Section 2.2, Section 3.2] The meta-analysis pools effect sizes across different operational definitions of creativity: creativity scales, originality scales, novelty scales, semantic distance, cosine similarity, and flexibility scores, rated by self, laypeople, experts, rule-based, or AI evaluators. With I2 ≈ 99% for RQ1, the pooled estimate is not interpretable as measuring a single latent construct unless the authors demonstrate that measurement type does not moderate the effects. The manuscript states that a subgroup analysis by construct found no substantial heterogeneity but omits it from the main text; please include these analyses or a measurement-type moderator in the paper or appendix.
  4. [Section 3.4] The RQ2b result (g = -0.86) is based on only 6 observations from 4 studies. Although the leave-one-out analysis is reassuring, the evidence base is small and the confidence interval is wide. The paper should more explicitly frame this finding as preliminary and should avoid definitive language such as 'consistent and statistically meaningful reduction' without acknowledging the small number of contributing studies.
minor comments (5)
  1. [Table 3] The table header contains the typo 'anaysis'; it should read 'analysis.'
  2. [Section 3.1] The paper reports 28 included studies but states that RQ1 uses 21 studies, RQ2a 12, and RQ2b 4; since some studies appear in multiple analyses, please clarify the overlap and the relation between the 28 studies and the 127 observation-level effect sizes.
  3. [Figure 1 and Section 2.1] The PRISMA flowchart and the eligibility criteria list different exclusion categories (e.g., 'insufficient study design' vs. 'insufficient statistics' vs. 'insufficient data reporting'), and the numbering in the text is inconsistent. Please align the flowchart, the numbered criteria, and the descriptions in the text.
  4. [Section 2.1] The screening and eligibility decisions were performed by one person (the first author). This is a potential source of bias and should be acknowledged as a limitation; PRISMA typically recommends dual screening.
  5. [Section 3.2 and Section 3.3] The phrases 'each study contributes less than 1.1% weight' and 'no single study contributes more than 5.1% weight' refer to per-observation weights, not study-level weights. Rephrase to avoid misleading readers into thinking each study is a single unit.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the meta-analytic estimates are standard aggregations of externally reported effect sizes.

full rationale

The paper is a systematic review and meta-analysis. Its headline results (g = -0.05, g = 0.27, g = -0.86) are weighted averages of effect sizes computed from statistics reported in 28 external studies, using standard formulas (Cohen's d and Hedges' g correction) and a DerSimonian-Laird random-effects model. No parameter is fitted to a subset of data and then relabeled as a prediction: the pooled estimates are descriptive summaries of the included effect sizes, not derived quantities that presuppose the conclusions. The only self-citations (e.g., Feuerriegel et al. for definitions of GenAI and NLP-based text analysis) appear as background references and are not load-bearing for the meta-analytic estimates. No uniqueness theorem, ansatz, or prior work by the same authors is invoked to force the choice of model or to forbid alternatives. The paper's robustness checks are standard meta-analytic diagnostics. Concern about non-independence caused by counting 52 observations from one study (HAI15) as separate units is a serious statistical validity issue, but it is not circularity: the estimates still reduce to the extracted input data, which is the normal and transparent behavior of a meta-analysis. The derivation chain is therefore self-contained with respect to the external evidence it aggregates.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The paper introduces no invented entities and relies on no hand-fitted free parameters. Its central estimates come from aggregating external data under standard meta-analytic models. The key assumptions are domain assumptions about construct comparability, search completeness, and the adequacy of bias corrections, all of which are explicitly invoked in Sections 2.1, 2.2, and 3.2.

assumptions (4)
  • standard math Random-effects meta-analysis with the DerSimonian-Laird estimator adequately models the between-study heterogeneity.
    Section 2.3 uses DerSimonian-Laird and the resulting very high I-squared values (above 93%) show the results are sensitive to this modeling choice; the assumption is standard but questionable when heterogeneity is extreme.
  • domain assumption The creativity constructs across studies (creativity scale, originality, novelty, semantic distance, flexibility, cosine similarity) are commensurable enough to pool into one Hedges' g.
    Section 2.2 describes different measurements and evaluators and acknowledges conceptual overlap; the comparability of these metrics is load-bearing for all three pooled estimates.
  • domain assumption The three-database, title-based search plus manual additions captures the relevant population of experiments.
    Section 2.1 defines the search query, which is title-only, and one person performed screening; the funnel asymmetry and trim-and-fill imputations suggest the search may have missed studies.
  • domain assumption Egger's test and trim-and-fill are valid diagnostics for publication bias in this literature.
    Section 3.2.2 uses these tools and the trim-and-fill result reverses the RQ1 conclusion; the validity of that imputation is assumed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Generative AI and Creativity: A Systematic Literature Review and Meta-Analysis." pith.science (2026). https://pith.science/paper/EZMAHNPG

@misc{pith2026250517241,
  author       = {Pith},
  title        = {Pith review of: Generative AI and Creativity: A Systematic Literature Review and Meta-Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EZMAHNPG}},
  note         = {Machine review of arXiv:2505.17241}
}
read the original abstract

Generative artificial intelligence (GenAI) is increasingly used to support a wide range of human tasks, yet empirical evidence on its effect on creativity remains scattered. Can GenAI generate ideas that are creative? To what extent can it support humans in generating ideas that are both creative and diverse? In this study, we conduct a meta-analysis to evaluate the effect of GenAI on the performance in creative tasks. For this, we first perform a systematic literature search, based on which we identify n = 28 relevant studies (m = 8214 participants) for inclusion in our meta-analysis. We then compute standardized effect sizes based on Hedges' g. We compare different outcomes: (i) how creative GenAI is; (ii) how creative humans augmented by GenAI are; and (iii) the diversity of ideas by humans augmented by GenAI. Our results show no significant difference in creative performance between GenAI and humans (g = -0.05), while humans collaborating with GenAI significantly outperform those working without assistance (g = 0.27). However, GenAI has a significant negative effect on the diversity of ideas for such collaborations between humans and GenAI (g = -0.86). We further analyze heterogeneity across different GenAI models (e.g., GPT-3.5, GPT-4), different tasks (e.g., creative writing, ideation, divergent thinking), and different participant populations (e.g., laypeople, business, academia). Overall, our results position GenAI as an augmentative tool that can support, rather than replace, human creativity-particularly in tasks benefiting from ideation support.

Figures

Figures reproduced from arXiv: 2505.17241 by the authors.

Figure 1
Figure 1. PRISMA flow chart. The task type reflects the notion that different creative tasks tap into distinct aspects of creativity (and thus different cognitive processes), such as compositional skills, imaginative reasoning, or divergent thinking (i.e., the ability to generate multiple, varied ideas in response to an open-ended prompt). The studies in our analysis employ a variety of creative tasks, each targeting differen… view at source ↗
Figure 2
Figure 2. Pooled effect comparing the creative performance of humans vs. GenAI (RQ1). The forest plot summarizes the Hedges’ 𝑔 effect sizes and 95% confidence intervals for a direct comparison between humans vs. GenAI (treatment: GenAI vs. control: human alone). Out of the 127 observations, 100 observations (participants 𝑚 = 4582) compare differences in creative performance between humans and GenAI, and are thus included in t… view at source ↗
Figure 3
Figure 3. Heterogeneity analysis for RQ1 (creative performance of humans vs. GenAI). Violin plots show the distribution of observation-level Hedges’ 𝑔 for the comparison in creative performance between human vs. GenAI conditions. The comparison is stratified by (a) GenAI model, (b) participant background, and (c) task type. Subgroup analyses are reported only for categories with a sufficient number of observations to support … view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Pooled effect of the benefit from Human-GenAI collaboration on creative performance (RQ2a). The forest plot summarizes the Hedges’ 𝑔 effect sizes and 95% confidence intervals (treatment: human-GenAI collaboration versus control: human alone). Out of the 127 observation…
Figure 5
Figure 5. Figure 5: Heterogeneity analysis for RQ2a (creative performance of humans+GenAI vs. humans only). Violin plots show the distribution of observation-level Hedges’ 𝑔 for the benefit in creative performance of human-GenAI collaboration over a human-only condition. The comparison is…
Figure 6
Figure 6. Figure 6: Pooled effects (RQ2b). Forest plot summarizing the Hedges’𝑔 effect sizes and 95% confidence intervals for creative diversity at the observation level (treatment: human-GenAI vs. control: Human alone), across six individual experiments from four studies. Out of the 127 …

Discussion (0). Sign in to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. OpenAlex reports about 11 citations worldwide. Full citation record

  1. Faster, Higher, Stronger? The Impact of GenAI on Knowledge Work Productivity - Evidence from the Field

    cs.HC 2026-07 conditional novelty 6.0 of 10

    GenAI support cut task time across all three knowledge-work tasks but improved output quality only for packaging and creation, while lowering quality and widening quality gaps in knowledge acquisition.

  2. AutoSynthesis: An agentic system for automated meta-analysis

    cs.AI 2026-07 conditional novelty 4.0 of 10

    An LLM agent pipeline can perform an end-to-end meta-analysis, but its claim of close agreement with expert meta-analyses rests on a single, co-authored benchmark and unreleased code.

Reference graph

Works this paper leans on

76 extracted references · 38 canonical work pages · cited by 2 Pith papers

  1. [1]

    Jaan Aru. 2025. Artificial intelligence and the internal processes of creativity.The Journal of Creative Behavior(2025). doi:10.1002/jocb.1530

    work page doi:10.1002/jocb.1530 2025

  2. [2]

    Joshua Ashkinaze, Julia Mendelsohn, Li Qiwei, Ceren Budak, and Eric Gilbert. 2024. How AI ideas affect the creativity, diversity, and evolution of human ideas: Evidence from a large, dynamic experiment. https://arxiv.org/abs/2401.13481

    arXiv 2024

  3. [3]

    Hedges, Julian P

    Michael Borenstein, Larry V. Hedges, Julian P. T. Higgins, and Hannah R. Rothstein. 2021.Introduction to meta-analysis. John Wiley & Sons, Ltd, West Sussex, UK. doi:10.1002/9780470743386

    doi:10.1002/9780470743386 2021

  4. [4]

    Boudreau and Karim R

    Kevin J. Boudreau and Karim R. Lakhani. 2013. Using the crowd as an innovation partner.Harvard Business Review91, 4 (2013), 60–69

    work page 2013

  5. [5]

    Lane, Miaomiao Zhang, Vladimir Jacimovic, and Karim R

    Léonard Boussioux, Jacqueline N. Lane, Miaomiao Zhang, Vladimir Jacimovic, and Karim R. Lakhani. 2024. The crowdless future? Generative AI and creative problem-solving.Organization Science35, 5 (2024), 1589–1607. doi:10. 1287/orsc.2023.18430

    arXiv 2024

  6. [6]

    Erik Brynjolfsson, Danielle Li, and Lindsey Raymond. 2025. Generative AI at work.The Quarterly Journal of Economics 140, 2 (2025), 889–942. doi:10.1093/qje/qjae044

    doi:10.1093/qje/qjae044 2025

  7. [7]

    Noah Castelo, Zsolt Katona, Peiyao Li, and Miklos Sarvary. 2024. How AI outperforms humans at creative idea generation.A vailable at SSRN 4751779(2024). doi:10.2139/ssrn.4751779

    work page doi:10.2139/ssrn.4751779 2024

  8. [8]

    Gary Charness and Daniela Grieco. 2024. Creativity and AI.A vailable at SSRN 4686415(2024). doi:10.2139/ssrn.4686415

    work page doi:10.2139/ssrn.4686415 2024

Show all 76 references
  1. [9]

    Zenan Chen and Jason Chan. 2024. Large language model in creative work: The role of collaboration modality and user expertise.Management Science70, 12 (2024), 9101–9117. doi:10.1287/mnsc.2023.03014

    2024

  2. [10]

    2013.Statistical power analysis for the behavioral sciences(2nd ed.)

    Jacob Cohen. 2013.Statistical power analysis for the behavioral sciences(2nd ed.). Routledge, New York. doi:10.4324/ 9780203771587

    2013

  3. [11]

    Zheyuan Kevin Cui, Mert Demirer, Sonia Jaffe, Leon Musolff, Sida Peng, and Tobias Salz. 2024. The effects of generative ai on high skilled work: Evidence from three field experiments with software developers.A vailable at SSRN 4945566 (2024). doi:10.2139/ssrn.4945566

    2024 doi

  4. [12]

    María-Isabel de Vicente-Yagüe-Jara, Olivia López-Martínez, Verónica Navarro-Navarro, and Francisco Cuéllar-Santiago

  5. [13]

    Deeks, Julian P

    Jonathan J. Deeks, Julian P. T. Higgins, Douglas G. Altman, and on behalf of the Cochrane Statistical Methods Group

  6. [14]

    Rebecca DerSimonian and Nan Laird. 1986. Meta-analysis in clinical trials.Controlled Clinical Trials7, 3 (1986), 177–188. doi:10.1016/0197-2456(86)90046-2

    1986 doi

  7. [15]

    Doshi, Sen Chai, and Matthias Troebinger

    Anil R. Doshi, Sen Chai, and Matthias Troebinger. 2024. How experience moderates the impact of generative AI ideas on the research process.A vailable at SSRN 5013086(2024). doi:10.2139/ssrn.5013086

    2024 doi

  8. [16]

    Doshi and Oliver P

    Anil R. Doshi and Oliver P. Hauser. 2024. Generative AI enhances individual creativity but reduces the collective diversity of novel content.Science Advances10, 28 (2024), eadn5290. doi:10.1126/sciadv.adn5290

    2024 doi

  9. [18]

    Sue Duval and Richard Tweedie. 2000. Trim and fill: a simple funnel-plot–based method of testing and adjusting for publication bias in meta-analysis.Biometrics56, 2 (2000), 455–463. doi:10.1111/j.0006-341x.2000.00455.x

    2000

  10. [19]

    Matthias Egger, George Davey Smith, Martin Schneider, and Christoph Minder. 1997. Bias in meta-analysis detected by a simple, graphical test.bmj315, 7109 (1997), 629–634. doi:10.1136/bmj.315.7109.629

    1997 doi

  11. [20]

    Anja Eisenreich, Julian Just, Daniela Gimenez-Jimenez, and Johann Füller. 2024. Revolution or inflated expectations? Exploring the impact of generative AI on ideation in a practical sustainability context.Technovation138 (2024), 103123. doi:10.1016/j.technovation.2024.103123

    2024

  12. [21]

    Stefan Feuerriegel, Jochen Hartmann, Christian Janiesch, and Patrick Zschech. 2024. Generative AI.Business & Information Systems Engineering66, 1 (2024), 111–126. doi:10.1007/s12599-023-00834-7

    2024 doi

  13. [22]

    Robertson, Steve Rathje, Jochen Hartmann, Saif M

    Stefan Feuerriegel, Abdurahman Maarouf, Dominik Bär, Dominique Geissler, Jonas Schweisthal, Nicolas Pröllochs, Claire E. Robertson, Steve Rathje, Jochen Hartmann, Saif M. Mohammad, Oded Netzer, Alexandra A. Siegel, Barbara Plank, and Jay J. van Bavel. 2025. Using natural langu...

    2025 doi

  14. [23]

    Carlos Gómez-Rodríguez and Paul Williams. 2023. A confederacy of models: A comprehensive evaluation of LLMs on creative writing. https://arxiv.org/abs/2310.08433v1

    2023 arXiv

  15. [24]

    Simone Grassini and Mika Koivisto. 2025. Artificial creativity? Evaluating AI against human performance in creative interpretation of visual stimuli.International Journal of Human–Computer Interaction41, 7 (2025), 4037–4048. doi:10. 1080/10447318.2024.2345430 Generative AI and...

    2025

  16. [25]

    Forward flow

    Kurt Gray, Stephen Anderson, Eric Evan Chen, John Michael Kelly, Michael S. Christian, John Patrick, Laura Huang, Yoed N. Kenett, and Kevin Lewis. 2019. “Forward flow”: A new measure to quantify free thought and predict creativity. American Psychologist74, 5 (2019), 539–554. d...

    2019 doi

  17. [26]

    Matthew Grimes, Georg von Krogh, Stefan Feuerriegel, Floor Rink, and Marc Gruber. 2023. From Scarcity to Abundance: Scholars and Scholarship in an Age of Generative Artificial Intelligence.Academy of Management Journal66, 6 (2023), 1617–1624. doi:10.5465/amj.2023.4006

    2023

  18. [27]

    Christensen, Philip R

    Joy Paul Guilford, Paul R. Christensen, Philip R. Merrifield, and Robert C. Wilson. 1978. Alternate uses. (1978). doi:10.1037/t06443-000

    1978 doi

  19. [28]

    Steffen Herbold, Annette Hautli-Janisz, Ute Heuer, Zlata Kikteva, and Alexander Trautsch. 2023. A large-scale comparison of human-written versus ChatGPT-generated essays.Scientific Reports13, 1 (2023), 18617. doi:10.1038/ s41598-023-45644-9

    2023

  20. [29]

    Hubert, Kim N

    Kent F. Hubert, Kim N. Awa, and Darya L. Zabelina. 2024. The current state of artificial intelligence generative language models is more creative than humans on divergent thinking tasks.Scientific Reports14, 1 (2024), 3440. doi:10.1038/s41598-024-53303-w

    2024 doi

  21. [30]

    Mete Ismayilzada, Debjit Paul, Antoine Bosselut, and Lonneke van der Plas. 2024. Creativity in AI: Progresses and challenges. https://arxiv.org/abs/2410.17218v4

    2024 arXiv

  22. [31]

    Mete Ismayilzada, Claire Stevenson, and Lonneke van der Plas. 2024. Evaluating creative short story generation in humans and large language models. https://arxiv.org/abs/2411.02316v5

    2024 arXiv

  23. [32]

    Dan Jackson and Jack Bowden. 2016. Confidence intervals for the between-study variance in random-effects meta- analysis using generalised heterogeneity statistics: should we use unequal tails?BMC Medical Research Methodology 16, 1 (2016), 118. doi:10.1186/s12874-016-0219-y

    2016 doi

  24. [33]

    Olson, Johnny Nahas, Denis Chmoulevitch, Simon J

    Jay A. Olson, Johnny Nahas, Denis Chmoulevitch, Simon J. Cropper, and Margaret E. Webb. 2021. Naming unrelated words predicts creativity.Proceedings of the National Academy of Sciences118, 25 (2021), e2022340118. doi:10.1073/ pnas.2022340118

    2021

  25. [34]

    Jennifer Haase and Paul H.P. Hanel. 2023. Artificial muses: Generative artificial intelligence chatbots have risen to human-level creativity.Journal of Creativity33, 3 (2023), 100066. doi:10.1016/j.yjoc.2023.100066

    2023

  26. [35]

    Plucker, Ronald A

    Jonathan A. Plucker, Ronald A. Beghetto, and Gayle T. Dow and. 2004. Why isn’t creativity more important to educational psychologists? Potentials, pitfalls, and future directions in creativity research.Educational Psychologist39, 2 (2004), 83–96. doi:10.1207/s15326985ep3902{_}1

    2004 doi

  27. [36]

    Anuj Kapoor and Madhav Kumar. 2025. Frontiers: Generative AI and personalized video advertisements.Marketing Science(2025). doi:10.1287/mksc.2023.0494

    2025

  28. [37]

    Mika Koivisto and Simone Grassini. 2023. Best humans still outperform artificial intelligence in a creative divergent thinking task.Scientific Reports13, 1 (2023), 13601. doi:10.1038/s41598-023-40858-3

    2023 doi

  29. [38]

    Larry V. Hedges. 1981. Distribution theory for glass’s estimator of effect size and related estimators.Journal of Educational Statistics6, 2 (1981), 107–128. doi:10.3102/10769986006002107

    1981 doi

  30. [39]

    Byung Cheol Lee and Jaeyeon Chung. 2024. An empirical investigation of the impact of ChatGPT on creativity.Nature Human Behaviour8, 10 (2024), 1906–1914. doi:10.1038/s41562-024-01953-1

    2024 doi

  31. [40]

    Runco and Garrett J

    Mark A. Runco and Garrett J. Jaeger and. 2012. The standard definition of creativity.Creativity Research Journal24, 1 (2012), 92–96. doi:10.1080/10400419.2012.650092

    2012

  32. [41]

    Jack McGuire, David de Cremer, and Tim van de Cruys. 2024. Establishing the importance of co-creation and self-efficacy in creative collaboration with artificial intelligence.Scientific Reports14, 1 (2024), 18525. doi:10.1038/s41598-024-69423-2

    2024 doi

  33. [42]

    Brewis, Fortune Nwaiwu, Deshan Sumanathilaka, Fernando Alva-Manchego, and Joanna Demaree-Cotton

    Peidong Mei, Deborah N. Brewis, Fortune Nwaiwu, Deshan Sumanathilaka, Fernando Alva-Manchego, and Joanna Demaree-Cotton. 2025. If ChatGPT can do it, where is my creativity? Generative AI boosts performance but diminishes experience in creative writing.Computers in Human Behavi...

    2025

  34. [43]

    Noah Bohren, Rustamdjan Hakimov, and Rafael Lalive. 2024. Creative and strategic capabilities of Generative AI: Evidence from large-scale experiments: IZA Discussion Papers. https://hdl.handle.net/10419/305744

    2024

  35. [44]

    Edenbaum, Joshua D

    William Orwig, Emma R. Edenbaum, Joshua D. Greene, and Daniel L. Schacter. 2024. The language of creativity: Evidence from humans and large language models.The Journal of Creative Behavior58, 1 (2024), 128–136. doi:10. 1002/jocb.636

    2024

  36. [45]

    Page, Joanne E

    Matthew J. Page, Joanne E. McKenzie, Patrick M. Bossuyt, Isabelle Boutron, Tammy C. Hoffmann, Cynthia D. Mulrow, Larissa Shamseer, Jennifer M. Tetzlaff, Elie A. Akl, Sally E. Brennan, Roger Chou, Julie Glanville, Jeremy M. Grimshaw, Asbjørn Hróbjartsson, Manoj M. Lalu, Tianjin...

  37. [46]

    Page, Joanne E

    Matthew J. Page, Joanne E. McKenzie, Patrick M. Bossuyt, Isabelle Boutron, Tammy C. Hoffmann, Cynthia D. Mulrow, Larissa Shamseer, Jennifer M. Tetzlaff, Elie A. Akl, Sally E. Brennan, Roger Chou, Julie Glanville, Jeremy M. Grimshaw, 20 Niklas Holzner, Sebastian Maier, and Stef...

  38. [47]

    Satu Parjanen. 2012. Experiencing creativity in the organization: From individual creativity to collective creativity. Interdisciplinary Journal of Information, Knowledge & Management7 (2012), 109–128. doi:10.28945/1580

    2012 doi

  39. [48]

    Peterson and Steven P

    Robert A. Peterson and Steven P. Brown. 2005. On the use of beta coefficients in meta-analysis.Journal of Applied Psychology90, 1 (2005), 175–181. doi:10.1037/0021-9010.90.1.175

    2005 doi

  40. [49]

    Rosnow and Robert Rosenthal

    Ralph L. Rosnow and Robert Rosenthal. 1996. Computing contrasts, effect sizes, and counternulls on other people’s published data: General procedures for research consumers.Psychological Methods1, 4 (1996), 331. doi:10.1037/1082- 989X.1.4.331

    1996 doi

  41. [50]

    doi:10.1136/bmj.n71

    PRISMA 2020 Flow Diagram. doi:10.1136/bmj.n71

    2020 doi

  42. [51]

    Solve Sæbø and Helge Brovold. 2024. On the stochastics of human and artificial creativity. https://arxiv.org/abs/2403. 06996v1

    2024

  43. [52]

    Max Schemmer, Patrick Hemmer, Maximilian Nitsche, Niklas Kühl, and Michael Vössing. 2022. A meta-analysis of the utility of explainable artificial intelligence in human-AI decision-making. InAAAI/ACM Conference on AI, Ethics, and Society (AIES). doi:10.1145/3514094.3534128

    2022

  44. [53]

    Max Schemmer, Niklas Kuehl, Carina Benz, Andrea Bartos, and Gerhard Satzger. 2023. Appropriate reliance on AI advice: Conceptualization and the effect of explanations. InInternational Conference on Intelligent User Interfaces (IUI). doi:10.1145/3581641.3584066

    2023

  45. [54]

    Suebsarn Ruksakulpiwat, Lalipat Phianhasin, Chitchanok Benjasirisan, Kedong Ding, Anuoluwapo Ajibade, Ayanesh Kumar, and Cassie Stewart. 2024. Assessing the efficacy of ChatGPT versus human researchers in identifying relevant studies on mHealth interventions for improving medi...

    2024 doi

  46. [55]

    Yu Song, Longchao Huang, Lanqin Zheng, Mengya Fan, and Zehao Liu. 2025. Interactions with generative AI chatbots: unveiling dialogic dynamics, students’ perceptions, and practical competencies in creative problem-solving. International Journal of Educational Technology in High...

    2025 doi

  47. [56]

    Liz Sanders SonicRim. 2001. Collective creativity.Design6, 3 (2001), 1–6. http://www.echo.iat.sfu.ca/library/sanders_ 01_collective_creativity.pdf

    2001

  48. [57]

    Jonathan A. C. Sterne, Jelena Savović, Matthew J. Page, Roy G. Elbers, Natalie S. Blencowe, Isabelle Boutron, Christo- pher J. Cates, Hung-Yuan Cheng, Mark S. Corbett, Sandra M. Eldridge, Jonathan R. Emberson, Miguel A. Hernán, Sally Hopewell, Asbjørn Hróbjartsson, Daniela R. ...

    2019

  49. [58]

    Chenglei Si, Diyi Yang, and Tatsunori Hashimoto. 2024. Can LLMs generate novel research ideas? A large-scale human study with 100+ NLP researchers. https://arxiv.org/abs/2409.04109

    2024 arXiv

  50. [59]

    Yuan Sun, Eunchae Jang, Fenglong Ma, and Ting Wang. 2024. Generative AI in the wild: Prospects, challenges, and strategies. InCHI Conference on Human Factors in Computing Systems (CHI). doi:10.1145/3613904.3642160

    2024

  51. [60]

    Paletz and Kaiping Peng

    Susannah B.F. Paletz and Kaiping Peng. 2008. Implicit theories of creativity across cultures: Novelty and appropriateness in two product domains.Journal of Cross-Cultural Psychology39, 3 (2008), 286–302. doi:10.1177/0022022108315112

    2008 doi

  52. [61]

    Michelle Vaccaro, Abdullah Almaatouq, and Thomas Malone. 2024. When combinations of humans and AI are useful: A systematic review and meta-analysis.Nature Human Behaviour8, 12 (2024), 2293–2303. doi:10.1038/s41562-024-02024-1

    2024 doi

  53. [62]

    Luning Sun, Yuzhuo Yuan, Yuan Yao, Yanyan Li, Hao Zhang, Xing Xie, Xiting Wang, Fang Luo, and David Stillwell

  54. [63]

    Wolfgang Viechtbauer and Mike W.-L. Cheung. 2010. Outlier and influence diagnostics for meta-analysis.Research Synthesis Methods1, 2 (2010), 112–125. doi:10.1002/jrsm.11

    2010 doi

  55. [64]

    Kelly, Saumya Pareek, Qiushi Zhou, and Eduardo Velloso

    Samangi Wadinambiarachchi, Ryan M. Kelly, Saumya Pareek, Qiushi Zhou, and Eduardo Velloso. 2024. The effects of generative AI on design fixation and divergent thinking. InCHI Conference on Human Factors in Computing Systems (CHI). doi:10.1145/3613904.3642919

    2024

  56. [65]

    Su, Zhun Deng, Michael Qizhe Xie, Hannah Brown, and Kenji Kawaguchi

    Haonan Wang, James Zou, Michael Mozer, Anirudh Goyal, Alex Lamb, Linjun Zhang, Weijie J. Su, Zhun Deng, Michael Qizhe Xie, Hannah Brown, and Kenji Kawaguchi. 2024. Can AI be as creative as humans? https://arxiv.org/ abs/2401.01623v4

    2024 arXiv

  57. [66]

    Emily Wenger and Yoed Kenett. 2025. We’re different, we’re the same: Creative homogeneity across LLMs. https: //arxiv.org/abs/2501.19361v1 Generative AI and Creativity: A Systematic Literature Review and Meta-Analysis 21

    2025 arXiv

  58. [67]

    Wolfgang Viechtbauer. 2010. Conducting meta-analyses in R with the metafor package.Journal of Statistical Software 36, 3 (2010), 1–48. doi:10.18637/jss.v036.i03

    2010 doi

  59. [68]

    Zhikun Wu, Thomas Weber, and Florian Müller. 2025. One does not simply meme alone: Evaluating co-creativity between LLMs and humans in the generation of humor. InInternational Conference on Intelligent User Interfaces (IUI). doi:10.1145/3708359.3712094

    2025

  60. [69]

    Cheng Xu, Shuhao Guan, Derek Greene, and M-Tahar Kechadi. 2024. Benchmark data contamination of large language models: A survey. https://arxiv.org/abs/2406.04244v1

    2024 arXiv

  61. [70]

    Man Zhang, Ying Li, Yang Peng, Yijia Sun, Wenxin Guo, Huiqing Hu, Shi Chen, and Qingbai Zhao. 2025. AI delivers creative output but struggles with thinking processes. https://arxiv.org/abs/2503.23327v1

    2025 arXiv

  62. [71]

    Jiexin Zheng, Ka Chau Wong, Jiali Zhou, and Tat Koon Koh. 2024. Large language model in ideation for product innovation: An exploratory comparative study.A vailable at SSRN 4729982(2024). doi:10.2139/ssrn.4729982

    2024 doi

  63. [72]

    Wilson, Joy P

    Robert C. Wilson, Joy P. Guilford, and Paul R. Christensen. 1953. The measurement of individual differences in originality.Psychological Bulletin50, 5 (1953), 362–370. doi:10.1037/h0060857

    1953 doi

  64. [73]

    Wenbo Zou and Feng Zhu. 2025. Generative AI adoption in human creative tasks: Experimental evidence.A vailable at SSRN 5196748(2025). doi:10.2139/ssrn.5196748 22 Niklas Holzner, Sebastian Maier, and Stefan Feuerriegel Acknowledgment of AI Usage & Overview of tools usedDuring t...

    2025 doi

  65. [77]

    Eric Zhou and Dokyun Lee. 2024. Generative artificial intelligence, human creativity, and art.PNAS Nexus3, 3 (2024), pgae052. doi:10.1093/pnasnexus/pgae052

    2024 doi

  66. [2019]

    InCochrane Handbook for Systematic Reviews of Interventions

    Analysing data and undertaking meta-analyses: 10. InCochrane Handbook for Systematic Reviews of Interventions. John Wiley & Sons, Ltd, 241–284. doi:10.1002/9781119536604.ch10

    doi

  67. [2023]

    doi:10.3916/C77-2023-04

    Writing, creativity, and artificial intelligence: ChatGPT in the university context.Comunicar: Media Education Research Journal31, 77 (2023), 45–54. doi:10.3916/C77-2023-04

    2023 doi

  68. [2024]

    https://arxiv.org/ abs/2412.03151

    Large language models show both individual and collective creativity comparable to humans. https://arxiv.org/ abs/2412.03151

    arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.