The DOI matches its Crossref record; 12 of the 12 references it cites were found in a registry or an index; 3 of the 4 sources its claims cite were checked for support.
Human-AI Co-Creation in Visual Communication Design: A Framework and Empirical Study
Open accessAbstract
Generative artificial intelligence is changing visual communication design from a tool-assisted activity into a collaborative process in which designers and AI systems share work across ideation, generation, evaluation, and refinement. Existing studies discuss AI support for image generation, prompting, and design exploration, but they offer limited guidance on how human creativity, design judgment, and AI-generated visual outputs can be organized within a complete design workflow. This paper proposes a human-AI co-creation framework for visual communication design. The framework contains four stages: design problem definition, AI-assisted visual exploration, human-led selection and refinement, and final design evaluation. It also clarifies human-led tasks, AI-supported tasks, and shared review tasks. An empirical study with professional visual designers was conducted to examine how AI participation affects design efficiency, creative exploration, visual quality, and designer experience. The results show that human-AI co-creation supports faster visual prototyping and wider idea generation, while human designers remain necessary for design intention, semantic accuracy, aesthetic judgment, and contextual fit. The findings are presented as trend-level evidence because the study used a modest sample and one poster-design brief, which limits generalizability. The study provides a structured framework and empirical evidence for understanding human-AI collaboration in intelligent visual communication design.
Overview
CoChat Summary
TL;DR
The paper proposes a four-stage human-AI co-creation framework for visual communication design (problem definition, AI-assisted visual exploration, human-led selection and refinement, final evaluation) and tests it in a controlled study with 30 professional designers. Results show AI sped up prototyping and broadened exploration while human designers remained necessary for semantic accuracy and contextual fit, though the authors frame all differences as trend-level evidence.
Key findings
- The Co-Creation group completed the task faster: total time fell 20.2%, from 114.5 to 91.4 minutes, with Ideation and Drafting dropping from 45.2 to 18.5 minutes while Refinement increased.
- AI expanded exploration: the Co-Creation group produced 2.56 times as many concepts as the Traditional group and scored higher on divergent thinking (a 2.2-point average gain); the concept discard rate in the Co-Creation group was 78%.
- Quality ratings (7-point Likert) favored co-creation on Aesthetic Appeal and Output Novelty, with overall quality rising from 5.30 to 6.23 (0.93-point gain), but Semantic Accuracy was slightly lower (a 0.2-point decrease).
- Aggregate trends reported: a 23.1-minute average time reduction, 5.3 more concepts, and a 0.93-point overall quality gain.
Methods
- Empirical study with 30 professional visual designers randomly divided into a Traditional Group (n=15, standard software such as Adobe Illustrator/Photoshop, no generative AI) and a Co-Creation Group (n=15, framework plus text-to-image AI tools); both received an identical brief to design a visual identity and promotional poster for a technology ethics conference, time-boxed to 120 minutes.
- Metrics covered design efficiency (time per phase), creative exploration (number of distinct concepts and expert-judged divergent thinking scores on a 1–10 scale), and subjective evaluation (expert ratings of visual quality, semantic accuracy, and novelty; participant satisfaction on a 1–7 Likert scale), with blind expert judging of final posters and intermediate files used to count concepts.
- Given the sample size and single poster scenario, analysis reports descriptive statistics only—group means, mean differences, relative changes, and concept expansion—without formal significance tests or variance measures.
Limitations
- Stated by the paper: the small experimental scale (30 designers) and single poster brief limit generalizability; quantitative differences are trend-level rather than strong statistical conclusions; the proposed mechanism of creative expansion and semantic drift is based on author interpretation and lacks process-level evidence such as prompt histories, revision traces, or stratified analyses by designer profile.
- Stated by the paper: future work should test more domains and experience levels and collect prompt histories, process logs, standard deviations, confidence intervals, and inferential test results.
- Evident from the text: no significance testing or variance measures were performed, so the reported group differences cannot be assessed for statistical reliability; the study also relies on a single 120-minute task with one brief type.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
1 Valid · 4 Not valid · 1 Not checked · 29 Own results and interpretations
At least one cited source doesn’t support this claim or points to a different paper, and none confirmed it — see the verdicts below.
In the paper
“This is consistent with prior findings that generative AI can reduce design fixation during early ideation [10].”
§ 5 Results > 5.2 Creative Exploration and Divergence
What the sources say
- Disputedcites [10] [10] S. Wadinambiarachchi, R. M. Kelly, S. Pareek, Q. Zhou, and E. Velloso. 2024. The Effects of Generative AI on Design Fixation and Divergent Thinking. In Proceedings of the CHI Conference on Human Factors in Computing Systems. Association for Computing Machinery, New York, NY, USA, 1–18. Digital Library Google Scholar a [...] careful study. Wadinambiarachchi et al. b [...] exploring divergent aesthetic directions c [...] design fixation during early ideation sourceExperimental — accuracy not yet measured
The Results (§4.6) say generative-AI support led to higher design fixation. The Conclusion (§6) likewise reports higher fixation among novice designers using an AI image generator, not a reduction during ideation. The paper discusses ways future AI tools might mitigate fixation, but does not report that effect as a finding. Page numbers were not provided. (LLM, conf=0.93)
Checked by a reasoning model — may be wrong.
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Disputed
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“which supports the need for human-led refinement because AI outputs may drift away from the intended message [7].”
§ 5 Results > 5.3 Quality and Subjective Evaluation
What the sources say
- Could not verifycites [7] [7] W. Li. 2024. A Study on Factors Influencing Designers’ Behavioral Intention in Using AI-Generated Content for Assisted Design: Perceived Anxiety, Perceived Risk, and UTAUT. International Journal of Human-Computer Interaction 41, 2 (2024), 1064–1077. Google Scholar a [...] AI-based manufacturing design tools. Li b [...] may drift away from the intended message sourceExperimental — accuracy not yet measured
The Introduction describes generating images from textual descriptions, while §2.2 discusses designers’ perceived risks, chiefly ethical, copyright, and privacy concerns. Neither passage says AI outputs may depart from an intended message. Searches for intended message, deviation from design intent, and outputs inconsistent with prompts found no supporting passage in our copy; page numbers were not available. (LLM, conf=0.75)
hedgedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Could not verify
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The overall quality score increased from 5.30 to 6.23.”
§ 5 Results > 5.3 Quality and Subjective Evaluation
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The results show that human-AI co-creation supports faster visual prototyping and wider idea generation, while human designers remain necessary for design intention, semantic accuracy, aesthetic judgment, and contextual fit.”
- Also stated in: § 5.1 Design Efficiency · plain
“the Co-Creation group required less total time than the Traditional group, with a 20.2% reduction from 114.5 mins to 91.4 mins.”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 5.1 Design Efficiency · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“However, Semantic Accuracy was slightly lower than in the Traditional group, which supports the need for human-led refinement because AI outputs may drift away from the intended message [7].”
- Also stated in: § 5.3 Quality and Subjective Evaluation · plain
“The only negative direction observed at the trend level was a 0.2-point decrease in semantic accuracy.”
§ 5 Results > 5.3 Quality and Subjective Evaluation
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 5.3 Quality and Subjective Evaluation · plain
15 more claims
Create a free account to see all 35 claims and every source verdict.
- Probably not a substantial claim. 8
- It may state two separate claims as one. 8
Citations
Cites
The Effects of Generative AI on Design Fixation and Divergent Thinking
Wadinambiarachchi, S., Kelly, R. M., Pareek, S., Zhou, Q., & Velloso, E. (2024) · Proceedings of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642919 · Cited by 320
Han, Y., Qiu, Z., Cheng, J., & Lc, R. (2024) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642133 · Cited by 183
Li, W. (2025) · International Journal of Human–Computer Interaction · doi:10.1080/10447318.2024.2310354 · Cited by 167
User Experience Design Professionals’ Perceptions of Generative Artificial Intelligence
Li, J., Cao, H., Lin, L., Hou, Y., Zhu, R., & El Ali, A. (2024) · Proceedings of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642114 · Cited by 163
Gmeiner, F., Yang, H., Yao, L., Holstein, K., & Martelaro, N. (2023) · Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3544548.3580999 · Cited by 130
Designing with AI: An Exploration of Co-Ideation with Image Generators
Chiou, L.-Y., Hung, P.-K., Liang, R.-H., & Wang, C.-T. (2023) · Proceedings of the 2023 ACM Designing Interactive Systems Conference · doi:10.1145/3563657.3596001 · Cited by 113
Wang, S.-Y., Su, W. L., Chen, S., Tsai, C.-Y., Misztal, M., Cheng, K. M., Lin, A., Chen, Y., & Chen, M. Y. (2024) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642901 · Cited by 49
Generative AI Futures: A Speculative Design Exploration
Lin, L., & Long, D. (2023) · Creativity and Cognition · doi:10.1145/3591196.3596616 · Cited by 41
AI-Assisted Causal Pathway Diagram for Human-Centered Design
Zhong, R., Shin, D.-H., Meza, R. D., Klasnja, P., Colusso, L. F., & Hsieh, G. (2024) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642179 · Cited by 28
Cao, Y., Huang, Y., Truong, A., Shin, H. V., & Xia, H. (2025) · Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713401 · Cited by 16
Metadata
- Authors
- Zhang, D.
- Affiliations
- Hefei Science and Technology College (CN)
- Corresponding author
- Zhang, Dongju
- ORCID
- Zhang, Dongju
- Venue
- Proceedings of the 2026 International Conference on Artificial Intelligence and Generative Design
- Publisher
- ACM
- Published
- 2026-05-15
- Pages
- 77-80
- OpenAlex
- W7204989371
- Cited by
- 0
- References
- 3
- FWCI
- 0.00
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- Locations
- International Conference on Artificial Intelligence and Generative Design (publishedVersion, cc-by-nc-nd)
- Topics
- Physical Sciences › Computer Science › Computer Graphics and Computer-Aided Design › Digital Media and Visual Art
- Physical Sciences › Engineering › Mechanical Engineering › Design Education and Practice
- Social Sciences › Arts and Humanities › Visual Arts and Performing Arts › Art, Technology, and Culture
- Keywords
- human-AI co-creation, visual communication design, generative AI, design workflow, image generation, design evaluation, designer experience, creative exploration, design intent, aesthetic judgment
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.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-30, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-30 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
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
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Human-AI Co-Creation in Visual Communication Design: A Framework and Empirical Study. https://app.cochat.ai/papers/10.1145/3830911.3830925
- The paper
- Zhang, D. (2026). Human-AI Co-Creation in Visual Communication Design: A Framework and Empirical Study. Proceedings of the 2026 International Conference on Artificial Intelligence and Generative Design. https://doi.org/10.1145/3830911.3830925
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.