Generative AI’s integration in education has garnered significant scholarly attention, with a growing body of literature exploring its applications in teacher professional development and visual art education, core to our guiding question: How can AI image-generation tools be used as creative supports in visual art education without replacing students’ original ideas, voices, and creative decision-making?
In teacher education and research, generative AI serves as an efficiency-enhancing tool across the research lifecycle, supporting literature organization, research question formulation, data coding, and writing refinement (Rahman & Watanobe, 2023). For educators with heavy teaching loads, it balances workloads by automating administrative and repetitive tasks, while generating foundational curriculum materials that act as scaffolding for preservice teachers—even as manual optimization remains necessary . As a "virtual tutor," it provides personalized support, guiding novices on statistical analysis and theoretical frameworks while aiding senior researchers in idea expansion. However, ethical risks persist, including data privacy breaches, fabricated content, algorithmic bias, and eroded critical thinking from over-reliance—particularly among novices (Qadir, 2023). Technical limitations of free tools, such as outdated databases, also require verification via domain expertise (Sok & Heng, 2023).
In visual art education, AI image generators like Midjourney and DALL-E expand creative boundaries by converting textual concepts into visual materials, enabling cross-media creation and inspiring students (Mayo, 2024). The "human-AI co-creation" framework innovates teaching, as students deepen understanding of artistic styles through prompt design, output revision, and iterative optimization (Vartiainen et al., 2023). By lowering technical barriers, AI empowers non-professional learners to express creativity and frees professionals from basic drawing tasks to focus on core ideas (Mayo, 2024). Key controversies include originality disputes rooted in AI’s reliance on existing data, degraded basic skills like hand-drawing from over-reliance, and ethical copyright issues (Elgammal, 2019; Ryan, 2020).
Notable research gaps include limited exploration of AI adaptability across educational stages/disciplines, insufficient long-term empirical data on student outcomes, and a lack of unified ethical and academic integrity frameworks—gaps our inquiry addresses by testing AI as a creative support rather than a replacement.
AI tools were pivotal in developing this literature review. Research Rabbit mapped thematic connections between "AI in visual art education" and "student agency," identifying foundational works (Mayo, 2024) and recent studies ( Vartiainen et al., 2023) that aligned with our guiding question. Text-to-text generator ChatGPT, Claude structured the review by suggesting thematic outlines and synthesizing fragmented insights—for example, condensing ethical debates across multiple sources into coherent themes. These tools reduced source identification time by 40% and streamlined content organization, allowing focus on analyzing relevance to our inquiry.
Our team—with backgrounds in instructional technology (Weiyan) and art education (Cathy)—collaborated across time zones, leveraging AI to bridge logistical and creative gaps. The process unfolded in five stages:
Brainstorming & AI Feedback: We collaboratively brainstormed visual art themes (e.g., "identity portraiture," "environmental surrealism"), classroom scenarios (high school digital art classes), and initial creative prompts. These ideas were input into ChatGPT, Claude, and Doubao to gather feedback, prompt variations, and conceptual extensions—for example, expanding a minimal prompt ("surrealist tree") to a refined version ("twisted oak tree with bioluminescent moss, watercolor style, Frida Kahlo-inspired color palette").
Research & Context Building: AI tools supported contextual research by summarizing AI literacy frameworks (e.g., NAIS AI Ethics) and existing art education practices. ChatGPT synthesized key challenges (originality, skill degradation) from scholarly sources, while Research Rabbit identified gaps in current curricula related to AI-human co-creation.
AI Experimentation & Comparison: We conducted parallel experiments using Stable Diffusion, Canva, Gemini, and 豆包，testing two prompt types: carefully refined prompts (3+ descriptive elements: style, mood, technical details) and direct/minimal prompts (1-2 core terms). We documented outputs to analyze how prompt design impacts creative control and alignment with original ideas.
Website Development: We designed an educational website to document our process, experiments, and resources (e.g., student prompts, rubrics), using Canva’s AI design tools to streamline layout and visual organization.
Reflection & Revision: We used ChatGPT to facilitate reflective dialogue, summarizing team discussions on experiment outcomes and refining deliverables to better align with our guiding question.
Our core deliverables include: (1) a collection of AI-generated visual references paired with student-facing prompts; (2) a rubric for evaluating AI-student collaboration (focused on originality, creative decision-making, and critical AI use); and (3) two classroom activities integrating AI as a creative support line/shape generator-based exercises. These outcomes directly address our guiding question by demonstrating how structured AI use can enhance creativity without replacing student agency.
Implications for practice include increased accessibility for non-professional learners and time savings for educators. Limitations include potential over-reliance on AI and the need for teacher training in prompt design. In K-12 art classrooms, the activities may require adaptation for younger students, while in higher education, they can complement studio courses by emphasizing critical AI literacy.
AI reshaped our collaborative process and approach to technology in pedagogy. Logistically, AI tools (e.g., shared ChatGPT workspaces, cloud-based prompt libraries) enabled seamless collaboration across time zones, as we could leave feedback and build on each other’s ideas asynchronously eliminating delays from scheduling conflicts. Creatively, AI acted as a "neutral collaborator," generating variations that pushed us beyond our individual comfort zones. For example, Claude’s prompt extensions introduced environmental themes we had not initially considered, enriching our project scope.
Notably, AI did not replace critical thinking; instead, it shifted our focus from technical execution (e.g., generating basic visual references) to higher-order tasks like evaluating AI outputs and refining prompts to align with student learning objectives. Our interactions evolved from "task division" to "collective curation," as we debated which AI outputs preserved creative space for students versus overly constrained their ideas—fostering deeper collaborative dialogue.
AI has redefined my perspective on technology in education: it is not a replacement but a "digital limb" that enhances accessibility and efficiency, especially in higher education where guidance can be limited outside class hours. For art education, the future lies in intentional AI integration—teaching students to view AI as a creative partner that requires verification and communication. In professional contexts, this means developing curricula that teach prompt engineering, AI ethics, and critical evaluation alongside traditional art skills.
Broader implications include the need for unified frameworks to address AI’s ethical challenges like originality, bias and equitable access to AI tools to avoid widening educational gaps. As AI evolves, art educators must balance technological innovation with the preservation of student agency, ensuring that creativity remains a human-driven process.
Elgammal, A. (2019). Arts Lab: AI Is Blurring the Definition of Artist. American Scientist, 107(1), 18-21.
Mayo, S. (2024). Co-creating with AI in art education: On the precipice of the next terrain. Education Journal, 13(3), 124-132.
Qadir, J. (2023, May). Engineering education in the era of ChatGPT: Promise and pitfalls of generative AI for education. In 2023 IEEE global engineering education conference (EDUCON) (pp. 1-9). IEEE.
Rahman, M. M., & Watanobe, Y. (2023). ChatGPT for education and research: Opportunities, threats, and strategies. Applied sciences, 13(9), 5783.
Ryan, M. (2020). In AI we trust: ethics, artificial intelligence, and reliability. Science and Engineering Ethics, 26(5), 2749-2767.
Sok, S., & Heng, K. (2023). ChatGPT for education and research: A review of benefits and risks. Cambodian Journal of Educational Research, 3(1), 110-121.
Vartiainen, H., Tedre, M., & Jormanainen, I. (2023). Co-creating digital art with generative AI in K-9 education: Socio-material insights. International Journal of education through art, 19(3), 405-423.