CoChat
Sign Up
Cited by 150Creativity in Education and NeuroscienceInnovative Human-Technology InteractionArt, Technology, and Culture

Establishing the importance of co-creation and self-efficacy in creative collaboration with artificial intelligence

Open access
McGuire, Jack, De Cremer, David, Van de Cruys, Tim·2024·Scientific Reports
DOI 10.1038/s41598-024-69423-2PMC PMC11316096
Open this paper in CoChat — free accountOpen PDF

Abstract

The emergence of generative AI technologies has led to an increasing number of people collaborating with AI to produce creative works. Across two experimental studies, in which we carefully designed and programmed state-of-the-art human-AI interfaces, we examine how the design of generative AI systems influences human creativity (poetry writing). First, we find that people were most creative when writing a poem on their own, compared to first receiving a poem generated by an AI system and using sophisticated tools to edit it (Study 1). Following this, we demonstrate that this creativity deficit dissipates when people co-create with-not edit-AI and establish creative self-efficacy as an important mechanism in this process (Study 2). Thus, our findings indicate that people must occupy the role of a co-creator, not an editor, to reap the benefits of generative AI in the production of creative works.

AI snapshot

Written from the paper’s full text.

TL;DR

Across two experiments, people were less creative when they edited an AI-generated poem than when writing alone, but this deficit disappeared when they co-created with AI line-by-line. Creative self-efficacy mediated the benefit of co-creation over editing.

Key findings

  • Study 1: Expert poets rated poems from the human-only condition as more creative (M = 18.23, SD = 4.94) than poems from the human–AI editor condition (M = 12.55, SD = 4.32, p < 0.001) and the AI-only condition (M = 9.45, SD = 2.00, p < 0.001); human–AI editor poems were also rated more creative than AI-only poems (p < 0.001). Self-reported creativity was higher in the human condition (M = 23.88, SD = 8.40) than in the human–AI condition (M = 19.13, SD = 10.40, p = 0.016, Cohen’s D = 0.47).
  • Study 2: Expert evaluations differed across conditions, F(2, 149) = 5.77, p = 0.004, partial η² = 0.07. Co-creator human–AI poems (M = 14.70, SD = 5.06) were rated significantly more creative than human–AI editor poems (M = 12.53, SD = 4.45, p = 0.026) and did not differ significantly from human-only poems (M = 15.74, SD = 4.98, p = 0.278).
  • Creative self-efficacy differed across conditions, F(2, 149) = 7.01, p = 0.001, partial η² = 0.09: human (M = 4.71, SD = 1.45) and co-creator human–AI (M = 4.62, SD = 1.43) were higher than human–AI editor (M = 3.74, SD = 1.43; p = 0.001 and p = 0.003, respectively).
  • Mediation: creative self-efficacy mediated the effect of co-creator vs. editor on expert evaluations of creativity, indirect effect B = 0.78, SE = 0.39, 95% CI [0.16, 1.68]. The indirect effect for co-creator vs. human was not significant, B = 0.09, SE = 0.29, 95% CI [−0.50, 0.78].

Methods

  • Two randomized online experiments using poetry writing. Study 1 recruited 101 Prolific participants; 6 failed an attention check, leaving 96 participants in the human and human–AI conditions, plus 50 AI-only generated poems. Study 2 recruited 152 Prolific participants with no attention-check failures.
  • Study 1 compared human-only writing, human–AI editing of an AI-generated poem, and AI-only poem generation. Study 2 compared human-only writing, human–AI editing, and a co-creator human–AI condition where participant and AI took turns writing one line each and the participant selected the poem’s topical words.
  • Creativity was assessed by 10 professional poets per study using the Consensual Assessment Technique on a 1–40 scale. Study 2 also measured creative self-efficacy with a three-item scale (α = 0.92) and tested mediation using PROCESS Model 4 with 10,000 bootstrapped samples.

Limitations

  • The paper states that the artificial experimental settings may limit generalizability to real-world creative processes, that long-term effects of co-creating vs. editing were not tested, that writer expertise was not examined as a boundary condition, and that more advanced generative AI systems were not tested.
  • Evident from the text: both studies rely on online Prolific samples and a single creative domain, poetry, which may constrain generalizability beyond these samples and tasks.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41598-024-69423-2

Cites

Statistics

Claims the paper makes

0 of 55 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    In Study 2, the creativity deficit observed in Study 1 dissipates when people co-create with AI rather than edit AI-generated output.

    “Following this, we demonstrate that this creativity deficit dissipates when people co-create with—not edit—AI and establish creative self-efficacy as an important mechanism in this process (Study 2).”
    • Also stated in: § Introduction · plain
      “we further find that this creativity deficit disappears when people are instead placed in the role of a co-creator and this effect is explained by greater reported levels of creative self-efficacy (Study 2).”
    Quote in the paperSubstantial claim
  • Findingown data

    In Study 1, participants in the human condition self-reported their poems as more creative than participants in the human-AI condition.

    “Independent samples t-tests reveal that participants in the human condition were significantly more likely to self-report their poems as more creative (M = 23.88, SD = 8.40), when compared to participants in the human–AI condition (M = 19.13, SD = 10.40, p = 0.016, Cohen’s D = 0.47).”
    Quote in the paperSubstantial claim
  • Findingown data

    In Study 1, expert evaluations of creativity differed significantly across the three experimental groups.

    “A one-way ANOVA revealed that the creativity of the poems differed significantly across the three experimental groups, F(2, 145) = 63.48, p < 0.001, partial η2 = 0.47 (see Table 1).”
    Quote in the paperSubstantial claim
  • Findingown data

    In Study 1, poems in the human-AI condition were regarded as more creative than poems in the AI condition.

    “Poems in the human condition (M = 18.23, SD = 4.94) were regarded as more creative than poems in the human-AI condition (M = 12.55, SD = 4.32; p < 0.001) the AI condition (M = 9.45, SD = 2.00; p < 0.001)”
    Quote in the paperSubstantial claim
  • Findingown data

    In Study 1, poems in the human condition were regarded as more creative than poems in the AI condition.

    “Poems in the human condition (M = 18.23, SD = 4.94) were regarded as more creative than poems in the human-AI condition (M = 12.55, SD = 4.32; p < 0.001) the AI condition (M = 9.45, SD = 2.00; p < 0.001)”
    Quote in the paperSubstantial claim
  • May combine two claims. 13
  • Classified as probably not a substantial claim. 29

  • May combine two claims.

    Premisecitation

    Although several studies have compared AI-generated creativity against a human benchmark, a surprisingly limited amount of work has examined the creative output of human-AI dyads and the collaborative processes that undergird this.

    “Furthermore, although several studies have compared AI generated creativity against a human benchmark8,16, a surprisingly limited amount of work has examined the creative output of human-AI dyads and the collaborative processes that undergird this.”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [8] Köbis, N. & Mossink, L. D. Artificial intelligence versus Maya Angelou: Experimental evidence that people cannot differentiate AI-generated from human-written poetry. Comput. Hum. Behav. 114, 106553 (2021). source

      Not checked in the free run.

    • cites [16] Boussioux, L., N Lane, J., Zhang, M., Jacimovic, V. & Lakhani, K. R. The Crowdless Future? How Generative AI is Shaping the Future of Human Crowdsourcing. The Crowdless Future. Available at SSRN (2023). source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    AI-generated creative writing is repetitive, follows a homogenous form and writing style, and lacks originality and novelty.

    “In addition, AI generated creative writing has been found to be repetitive, following a homogenous form and writing style12,13, further contributing to its lack of originality and novelty14.”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [12] Padmakumar, V. & He, H. Does writing with language models reduce content diversity? Preprint at http://arXiv.org/2309.05196 (2023).

      Not checked in the free run.

    • cites [13] Perez, E. et al. Discovering language model behaviors with model-written evaluations. Preprint at http://arXiv.org/2212.09251 (2022).

      Not checked in the free run.

    • cites [14] Mukherjee, A. & Chang, H. Managing the creative frontier of generative AI: The novelty-usefulness tradeoff. Calif. Manag. Rev. 1, 1 (2023).

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductcitation

    In Study 1, poets were given the poems in different, randomly generated orders.

    “In line with creativity research that utilizes the CAT35–37, poets were given the poems in different, randomly generated orders and asked to rate the creativity of each poem on a 1 (lowest creativity score) to 40 (highest creativity score) scale.”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [35] Amabile, T. M. Motivation and creativity: Effects of motivational orientation on creative writers. J. Person. Soc. Psychol. 48(2), 393 (1985). source

      Not checked in the free run.

    • cites [36] Amabile, T. M., Hennessey, B. A. & Grossman, B. S. Social influences on creativity: The effects of contracted-for reward. J. Person. Soc. Psychol. 50(1), 14 (1986). source

      Not checked in the free run.

    • cites [37] Kaufman, J. C., Baer, J. & Cole, J. C. Expertise, domains, and the consensual assessment technique. J. Creat. Behav. 43(4), 223–233 (2009). source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    In Study 1, poets were not provided any information about the authors behind the poems or the circumstances under which they were written.

    “Importantly, poets were not provided any information about the authors behind the poems or the circumstances under which they were written.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    The poetry generation system used in Study 1 is a state-of-the-art, scientifically validated neural poetry generation system that has been shown to generate output comparable to human-written poetry.

    “The algorithm used to generate novel poetry is a state of the art, scientifically validated, neural poetry generation system that has been shown to generate output comparable to human-written poetry33.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [33] Van de Cruys, T. Automatic poetry generation from prosaic text. In Proc. 58th Annual Meeting of the Association for Computational Linguistics 2471–2480 (2020). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    In Study 1, six participants failed to correctly answer an instrumental attention check and were removed from subsequent analyses.

    “Six participants failed to correctly answer an instrumental attention check and were removed from subsequent analyses (see below).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

    Premisecitation

    Poetry embodies the most prototypical expression of creativity-spanning centuries of human history-and enables testing the effect of being a co-creator vs. editor on creative outcomes.

    “Finally, as poetry embodies the most prototypical expression of creativity—spanning centuries of human history31,32—and enables us to test the effect of being a co-creator vs. editor on creative outcomes, this research focuses on poetry writing as the medium of creative expression.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [31] Shelley, P. B. & Brett-Smith, H. F. B. A defence of Poetry 79 (Brodie, 1980).

      Not checked in the free run.

    • cites [32] Perkins, D. A history of Modern Poetry Vol. 1 (Harvard University Press, 1976).

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductown data

    In Study 1, the study used the Consensual Assessment Technique (CAT) to evaluate creativity.

    “To evaluate the creativity of participant’s poems, we utilized the Consensual Assessment Technique (CAT), one of the most highly regarded assessment tools in creativity research17.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [17] Amabile, T. M. Social psychology of creativity: A consensual assessment technique. J. Person. Soc. Psychol. 43, 5 (1982). source

      Not checked in the free run.

  • May combine two claims.

    Premiseown data

    Evaluations of the quality of poems produced by this poetry generation system revealed that raters considered the poems to be comparable to works by well-established English poets and superior to previous poetry systems across dimensions of fluency, coherence, meaningfulness, and poeticism.

    “Evaluations of the quality of poems produced by this poetry generation system revealed that raters considered the poems to be comparable to works by well-established English poets (W.H. Auden, E.E. Cummings, Philip Larkin, Sarojini Naidu, and Sylvia Plath), and superior to previous poetry systems (Hafez and Deep-speare) across dimensions of fluency, coherence, meaningfulness, and poeticism.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [33] Van de Cruys, T. Automatic poetry generation from prosaic text. In Proc. 58th Annual Meeting of the Association for Computational Linguistics 2471–2480 (2020). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductnone given

    All conditions, measures and exclusions are reported; data are available at https://osf.io/3jqga/.

    “All conditions, measures and exclusions are reported; data are available at https://osf.io/3jqga/.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    The evolving nature of publicly available LLMs presents challenges because identical prompts at different points in time can produce very different results and changes to model specifications can lead to fluctuations in performance, even across short periods of time.

    “Moreover, the evolving nature of publicly available LLMs presents further challenges because identical prompts at different points in time can produce very different results20 and changes to model specifications can lead to fluctuations in performance, even across short periods of time21.”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [20] Aiyappa, R., An, J., Kwak, H., & Ahn, Y. Y. (2023). Can we trust the evaluation on ChatGPT? Preprint at http://arXiv.org/2303.12767. source

      Not checked in the free run.

    • cites [21] Chen, L., Zaharia, M. & Zou, J. How is ChatGPT's behavior changing over time? Preprint at http://arXiv.org/2307.09009 (2023).

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Such facets of co-creation will likely nurture intrinsic motivation, which is constructive for the development of self-efficacy.

    “Such facets of co-creation will likely nurture intrinsic motivation26, which is constructive for the development of self-efficacy30.”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [26] Roskes, M. Constraints that help or hinder creative performance: A motivational approach. Creat. Innov. Manag. 24(2), 197–206 (2015). source

      Not checked in the free run.

    • cites [30] Bandura, A. & Schunk, D. H. Cultivating competence, self-efficacy, and intrinsic interest through proximal self-motivation. J. Person. Soc. Psychol. 41(3), 586 (1981). source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductown data

    In Study 1, participants were randomly allocated to experimental condition groups.

    “Participants were randomly allocated to one of two experimental condition groups.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    Both experiments were approved for use with human subjects by an institutional review board.

    “Both experiments were approved for use with human subjects by an institutional review board.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductown data

    In Study 1, evaluations were obtained from 10 poets who are professional writers with published poems, and 6 possess a graduate degree in creative writing, English literature, or a related field.

    “We obtained evaluations from 10 poets in exchange for USD40 each. All poets we obtained evaluations from are professional writers and have had their poems published in either their own poetry books or as part of a poetry anthology that contains works from different poets. In addition, 6 of the poets possess a graduate degree (MA, MFA, PhD) in creative writing, English literature, or a related”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductnone given

    Condition assignments were random in both studies, with randomization administered via front-end programming.

    “Condition assignments were random in both studies, with randomization administered via front-end programming.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation

    The controlled experimental conditions do not fully capture the complexities of real-world creative processes where additional factors such as task type and organizational climate can influence creativity.

    “The controlled experimental conditions do not fully capture the complexities of real-world creative processes where additional factors such as task type and organizational climate62 can influence creativity.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [62] Mutonyi, B. R., Slåtten, T. & Lien, G. Organizational climate and creative performance in the public sector. Eur. Bus. Rev. 32(4), 615–631 (2020). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    The artificial settings of the experiments may affect the generalizability of the results to real-world environments.

    “First, the artificial settings of the experiments may affect the generalizability of the results to real-world environments.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    Our research does not address whether using a more advanced generative AI system would unearth different results.

    “Finally, our research does not address whether using a more advanced generative AI system would unearth different results.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data

    For AI to successfully augment human creativity, it is a requirement that it promotes creative self-efficacy and places humans in the role of a co-creator, not an editor.

    “Together, our findings suggest that for AI to successfully augment human creativity, it is a requirement that it promotes creative self-efficacy and places humans in the role of a co-creator, not an editor.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Co-creating (vs. editing) should elicit positive effects for creative self-efficacy and outcomes irrespective of the AI system in question.

  • Classified as probably not a substantial claim.

    Study design and conductown data

    In Study 2, participants were randomly allocated to one of three experimental condition groups.

    “Participants were randomly allocated to one of three experimental condition groups.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Prolific is an online platform explicitly designed for online participant recruitment by the scientific community and has been empirically demonstrated to provide higher quality data than alternative online platforms.

    “ProA is an online platform explicitly designed for online participant recruitment by the scientific community65 and has been empirically demonstrated to provide higher quality data than alternative online platforms66.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [65] Palan, S. & Schitter, C. Prolific.ac—A subject pool for online experiments. J. Behav. Exp. Financ. 17, 22–27 (2018). source

      Not checked in the free run.

    • cites [66] Peer, E., Brandimarte, L., Samat, S. & Acquisti, A. Beyond the Turk: Alternative platforms for crowdsourcing behavioral research. J. Exp. Soc. Psychol. 70, 153–163 (2017). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    In Study 2, the study used the Consensual Assessment Technique (CAT) to evaluate creativity, with ratings from 10 different poets.

    “To evaluate the creativity of participant’s poems, we utilized the Consensual Assessment Technique (CAT), one of the most highly regarded assessment tools in creativity research17. As in Study 1, we obtained ratings from 10 (different) poets in exchange for USD40 each.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [17] Amabile, T. M. Social psychology of creativity: A consensual assessment technique. J. Person. Soc. Psychol. 43, 5 (1982). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    In Study 2, no participants failed to correctly answer an instrumental attention check.

    “No participants failed to correctly answer an instrumental attention check.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    Data and syntax files are available on the Open Science Framework at https://osf.io/3jqga/.

    “Data and syntax files are available on the Open Science Framework at https://osf.io/3jqga/.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductcitation

    In Study 2, creative self-efficacy was measured using a three-item (α = 0.92) scale developed by Tierney and Farmer.

    “To measure creative self-efficacy, we adopted a three-item (α = 0.92) creative self-efficacy scale developed by Tierney and Farmer28.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [28] Tierney, P. & Farmer, S. M. Creative self-efficacy: Its potential antecedents and relationship to creative performance. Acad. Manag. J. 45(6), 1137–1148 (2002). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    Only in Study 1 did any participants fail the attention check (n = 6).

    “However, only in Study 1 did any participants fail this check (n = 6).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    All research studies received ethical approval from the Institutional Review Board (IRB) at the National University of Singapore (Ethical Approval Code: BIZ-MNO-20-0213).

    “All research studies received ethical approval from the Institutional Review Board (IRB) at the National University of Singapore (Ethical Approval Code: BIZ-MNO-20-0213)”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The long-term impact of co-creating (vs. editing) is not tested in this research.

    “A further methodological limitation of this research is the long-term impact of co-creating (vs. editing) is not tested.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    In both studies, we utilized attention checks.

    “Moreover, in both studies, we utilized attention checks”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductreasoning

    All methods of the reported studies were performed in accordance with the ethical guidelines and regulations of the IRB committee.

    “all methods of the reported studies were performed in accordance with the ethical guidelines and regulations of this committee”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    All manipulations and measures are reported.

    “All manipulations and measures are reported.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The expertise level of the human writer was not considered as a potential boundary condition in this research.

    “Additionally, the expertise level of the human writer was not considered as a potential boundary condition in this research.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Attention checks are an important method for enhancing data quality, particularly when utilizing online platforms such as Prolific.

    “an important method for enhancing data quality67, particularly when utilizing online platforms such as Prolific66.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [67] Oppenheimer, D. M., Meyvis, T. & Davidenko, N. Instructional manipulation checks: Detecting satisficing to increase statistical power. J. Exp. Soc. Psychol. 45(4), 867–872 (2009). source

      Not checked in the free run.

    • cites [66] Peer, E., Brandimarte, L., Samat, S. & Acquisti, A. Beyond the Turk: Alternative platforms for crowdsourcing behavioral research. J. Exp. Soc. Psychol. 70, 153–163 (2017). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductcitation

    We selected participants with an approval rating of at least 97%.

    “Following prior research, we selected participants with an approval rating of at least 97%.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    Informed consent was obtained for all participants.

    “Informed consent was obtained for all participants.”
    Quote in the paperNot a substantial claim
36 of 36

  • The authors state this as an assumption.

    The way people view their role in creative collaborations has an influence on the outcome of their endeavours.

    “We assume that the way people view their role in creative collaborations has an influence on the outcome of their endeavours22.”

  • The authors state this as an assumption.

    Co-creating (vs. editing) should elicit positive effects for creative self-efficacy and outcomes irrespective of the AI system in question.

    “Although we argue that co-creating (vs. editing) should elicit positive effects for creative self-efficacy and outcomes irrespective of the AI system in question, this assumption does warrant empirical validation.”
2 of 2

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Human collaboration still holds unique value in terms of emotional support, creative motivation, and shared sense of accomplishment.

    “Although AI has demonstrated significant advantages in many areas, human collaboration still holds unique value in terms of emotional support, creative motivation, and shared sense of accomplishment (Tang et al., 2024; McGuire et al., 2024).”

    From From humans to AI: understanding why AI is perceived as the preferred co-creation partner (2025) · cites this paper as McGuire et al., 2024

    The Introduction (page unavailable) suggests that giving people a co-creator role in human–AI work may nurture intrinsic motivation, and the Discussion says that role supports creative self-efficacy. The paper does not establish that collaboration between humans has unique value over AI collaboration, or address emotional support or a shared sense of accomplishment. (LLM, conf=0.84)

    Experimental — accuracy not yet measured

Statistics recomputed

Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.

  • ConsistentF(2, 145) = 63.48reported p < .001 · recomputed p < .001

    “F(2, 145) = 63.48, p < .001”

  • ConsistentF(2, 149) = 7.01reported p = .001 · recomputed p = .001

    “F(2, 149) = 7.01, p = .001”

  • ConsistentF(2, 149) = 6.08reported p = .003 · recomputed p = .003

    “F(2, 149) = 6.08, p = .003”

  • ConsistentF(2, 149) = 5.77reported p = .004 · recomputed p = .004

    “F(2, 149) = 5.77, p = .004”

4 of 4

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Study 2 human-condition expert-rating SD differs from Table 2

    Minor

    The results text gives an SD of 9.24 for the human condition's expert evaluations, whereas Table 2 gives 4.98 for the same measure and condition.

1 of 1

The citation lists haven’t been fetched yet

References 56 · Cited by 150

Research works printed but not linked: 12

  1. Jo, A. The promise and peril of generative AI. Nature 614(1), 214–216 (2023).
  2. Gozalo-Brizuela, R. & Garrido-Merchan, E. C. ChatGPT is not all you need. A state of the art review of large generative AI models. Preprint at http://arXiv.org/2301.04655 (2023).
  3. Anantrasirichai, N. & Bull, D. Artificial intelligence in the creative industries: A review. Artif. Intell. Rev. 1, 1–68 (2022).
  4. Fui-Hoon Nah, F., Zheng, R., Cai, J., Siau, K. & Chen, L. Generative AI and ChatGPT: Applications, challenges, and AI-human collaboration. J. Inf. Technol. Case Appl. Res. 25(3), 277–304 (2023).
  5. Padmakumar, V. & He, H. Does writing with language models reduce content diversity? Preprint at http://arXiv.org/2309.05196 (2023).
  6. Mukherjee, A. & Chang, H. Managing the creative frontier of generative AI: The novelty-usefulness tradeoff. Calif. Manag. Rev. 1, 1 (2023).
  7. Chen, L., Zaharia, M. & Zou, J. How is ChatGPT's behavior changing over time? Preprint at http://arXiv.org/2307.09009 (2023).
  8. Mirbabaie, M., Brünker, F., Möllmann, N. R. & Stieglitz, S. The rise of artificial intelligence–understanding the AI identity threat at the workplace. Electron. Mark. 1, 1–27 (2022).
  9. Shelley, P. B. & Brett-Smith, H. F. B. A defence of Poetry 79 (Brodie, 1980).
  10. Perkins, D. A history of Modern Poetry Vol. 1 (Harvard University Press, 1976).
  11. Hennessey, B. A. Self-determination theory and the social psychology of creativity. Psychol. Inq. 11(4), 293–298 (2000).
  12. Lai, M. C. & Chen, Y. C. Self-efficacy, effort, job performance, job satisfaction, and turnover intention: The effect of personal characteristics on organization performance. Int. J. Innov. Manag. Technol. 3(4), 387 (2012).

Authors
McGuire, J., De Cremer, D., & Van de Cruys, T.
Affiliations
Northeastern University (US)
Corresponding author
McGuire, Jack
ORCID
McGuire, Jack
Venue
Scientific Reports
Publisher
Springer Science and Business Media LLC
Published
2024-08-09
Volume
14
Issue
1
Pages
18525-18525
ISSN
2045-2322 (ISSN-L)
DOI
10.1038/s41598-024-69423-2
PMC
PMC11316096
OpenAlex
W4401454503
Cited by
150
References
56
Citations by year
2026: 90 · 2025: 54 · 2024: 6
FWCI
61.03
Type
journal article
Language
en
Open access
Licence
Locations
  • Scientific Reports (publishedVersion, cc-by-nc-nd)
  • PubMed Central (publishedVersion, cc-by-nc-nd)
  • PubMed (publishedVersion)
  • Lirias (KU Leuven) (submittedVersion, cc-by-nc-nd)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
Ministry of Education - Singapore (11/2019/115)
Topics
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Creativity in Education and Neuroscience
  • Physical Sciences › Computer Science › Human-Computer Interaction › Innovative Human-Technology Interaction
  • Social Sciences › Arts and Humanities › Visual Arts and Performing Arts › Art, Technology, and Culture
Keywords
human-AI collaboration, co-creation, creative self-efficacy, generative AI, creativity, poetry writing, human-AI interface
MeSH terms
Adult, Artificial Intelligence, Cooperative Behavior, Creativity, Female, Humans, Male, Poetry as Topic, Self Efficacy, Young Adult

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-09.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-09.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review