CoChat
Sign Up
Cited by 38Creativity in Education and NeuroscienceInnovative Human-Technology InteractionDigital Media and Visual Art

Agency in Human-AI Collaboration for Image Generation and Creative Writing: Preliminary Insights from Think-Aloud Protocols

Open access
Rafner, Janet, Zana, Blanka, Bang Hansen, Ida, Ceh, Simon, Sherson, Jacob, Benedek, Mathias, Lebuda, Izabela·2025·Creativity Research Journal
DOI 10.1080/10400419.2025.2587803
Open this paper in CoChat — free accountOpen PDF

Abstract

The integration of generative artificial intelligence into creative domains raises critical questions about human agency in AI-assisted creation. In this two-study article (study one, n = 6; study two, n = 7), we explored how people experience creative agency during co-creative tasks with AI, focusing on image generation and writing. Using think-aloud protocols and post-task semi-structured interviews, we conducted an analysis to identify four central dimensions of agency: creative self-efficacy, control over creative action, autonomy in the creative process, and ownership of the creative product, as well as the self-regulatory and metacognitive mechanisms that support sustaining agency.Preliminary findings reveal that agency in human–AI co-creation fluctuates across the creative process. Participants employed adaptive strategies—such as progressive refinement, selective appropriation, and counter-inspiration—to reinforce agency when navigating unpredictable AI outputs. However, moments of diminished control and uncertainty about ownership were common, particularly when AI contributions deviated from user expectations. To contextualize these experiences, we applied the Co-Creative Framework for Interaction Design, highlighting how tool affordances shaped participants’ sense of agency. Participants’ shifting experiences of agency suggest a need for adaptive system features that align with evolving user goals.

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1080/10400419.2025.2587803

Cites

122 of 124 references verified0 citation errors · 2 not checkable

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

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?

  • Meaningful revision requires the student to retain control over action, decision autonomy, and ownership of the final artefact.

    “The finding also connects to human-AI agency research because meaningful revision requires the student to retain control over action, decision autonomy, and ownership of the final artefact (Rafner et al. 2025).”

    From A Study on the Role of Midjourney, A Generative Artificial Intelligence Tool, in Enhancing Student Creativity in Illustration Design Classes (2026) · cites this paper as Rafner et al. 2025

    The sections “Is this mAIne?” and “Creativity as agentic action in human-AI co-creation” link creative control, autonomy, and ownership; participants who retained editorial control reported a stronger sense of ownership. The paper does not establish that retaining all three is required for meaningful revision by students. Its findings concern exploratory human-AI image generation and creative writing, not student revision specifically; page numbers were unavailable in our copy. (LLM, conf=0.88)

    Experimental — accuracy not yet measured

  • Creative control in human-AI co-creative tasks shifts according to action control, decision autonomy, creative self-efficacy, and perceived ownership.

    “Human-AI creative agency research further shows that creative control shifts across co-creative tasks according to action control, decision autonomy, creative self-efficacy, and perceived ownership of the final product (Rafner et al. 2025).”

    From A Study on the Role of Midjourney, A Generative Artificial Intelligence Tool, in Enhancing Student Creativity in Illustration Design Classes (2026) · cites this paper as Rafner et al. 2025

Self-check

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

The self-check needs the paper’s full text.

Cites

Flagged first, then by citations

Research works printed but not linked: 8

  1. Calderwood, A., Qiu, V., Gero, K. I., & Chilton, L. B. (2020,). How novelists use generative language models: An explora- tory user study. In HAI-GEN+ user2agent@ IUI. https:// ceur-ws.org/Vol-2848/HAI-GEN-Paper-3.pdf
  2. Efland, A. D. (1990). A history of art education: Intellectual and social currents in teaching the visual arts. Teachers College Press.
  3. Gero, J. S., & Kannengiesser, U. (2012). The situated function-behavior-structure framework. Design Studies, 33 (2), 111-133.
  4. Kantosalo, A., & Takala, T. (2020, September). Five c's for human-computer co-creativity-an update on classical crea- tivity perspectives. In Iccc (pp. 17-24). https://computatio nalcreativity.net/iccc20/papers/014-iccc20.pdf
  5. Morgan Kaufmann.
  6. Rafner, J., Zana, B., Beolet, T., Buyukguzel, S., Michel, E., Maiden, N., Risi, S., & Sherson, J. (2023). Crea. visions: A platform for casual co-creation with a purpose envision- ing the future through human-AI collaboration with multi- ple stakeholders. In A. Pease, J. M. Cunha, M. Ackerman, & D. G. Brown (Eds.) In Proceedings of the Fourteenth International Conference on Computational Creativity, ICCC, 2023 (pp. 293-297). Association for Computational Creativity. https://computationalcreativity.net/iccc23/ papers/ICCC-2023_paper_74.pdf
  7. Rogers, Y., Sharp, H., & Preece, J. (2011). Interaction design: Beyond human-computer interaction (3rd ed). Wiley.
  8. Sawyer, R. K. (2012). Explaining creativity: The science of human innovation (2nd ed.). Oxford University Press.

Cited by

Sorted by citations

Related

Authors
Rafner, J., Zana, B., Bang Hansen, I., Ceh, S., Sherson, J., Benedek, M., & Lebuda, I.
Affiliations
  • Aarhus University (DK)
  • Aarhus University
  • University of Wrocław (PL)
Corresponding author
Rafner, Janet
ORCID
Venue
Creativity Research Journal
Publisher
Informa UK Limited
Published
2025-12
Pages
1-24
ISSN
  • 1040-0419 (ISSN-L)
  • 1532-6934
DOI
10.1080/10400419.2025.2587803
OpenAlex
W7108203550
Cited by
38
References
97
Citations by year
2026: 36 · 2025: 2
FWCI
39.80
Type
journal article
Language
en
Open access
Licence
Locations
  • Creativity Research Journal (publishedVersion, cc-by)
  • OSF Preprints (OSF Preprints) (submittedVersion, cc-by)
Funders
  • SHAPE - Center for Shaping Digital Citizenship
  • Aarhus Institute of Advanced Studies (2024/54/E/HS6/00123)
  • Aarhus Universitets Forskningsfond (AUFF-E-2022-9-18)
  • Narodowe Centrum Nauki
Topics
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Creativity in Education and Neuroscience
  • Physical Sciences › Computer Science › Human-Computer Interaction › Innovative Human-Technology Interaction
  • Physical Sciences › Computer Science › Computer Graphics and Computer-Aided Design › Digital Media and Visual Art
Keywords
human-AI co-creation, creative agency, generative AI, image generation, creative writing, creative self-efficacy, creative autonomy, metacognition, self-regulation, progressive refinement, selective appropriation, tool affordances

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.

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