Incorporating generative artificial intelligence (GAI), such as DALL-E 2, as a tool in art education can provide an effective means for art educators to teach students to critically analyze art and enable students with little or no technical or artistic skills to participate in the creation of visual art, a new Rutgers study has found.

“GAI can provide art educators with an opportunity to engage students actively in the critical analysis of art or art-like material generated via AI through an experience-based approach, when educational experience leads to student discovery or construction of knowledge through activation of the imagination,” lead author SC&I Professor of Journalism and Media Studies John Pavlik said.

The study, “Art education and generative ai: An exploratory study in constructivist learning and visualization automation for the classroom,” co-authored with Orianna M. Pavlik, was published in the journal Creative Education.

The study’s findings are particularly important and relevant in this current educational environment, Pavlik said, because “educators face a dilemma regarding whether to embrace GAI as an opportunity for innovation in art education, or to reject its utilization as a threat to human artistic creativity or intellectual property rights.”

Pavlik said he and his coauthor undertook this research to learn more about how visual representations created via generative artificial intelligence (GenAI) could be utilized in teaching and learning and to do so ethically. 

“Many students have little skill in creating visualizations,” Pavlik said, “but they have great interest in this arena. GenAI presents an opportunity but is not without its limitations. A study such as this helps illuminate both the potential and the pitfalls.”

To conduct the study, Pavlik said they designed and conducted a qualitative exploratory analysis using the generative AI visual content platform DALL-E 2 (which is the second generation of DALL-E) which uses ChatGPT, a text-based large language model also developed by OpenAI.

Pavlik said he and his coauthor undertook this research to learn more about how visual representations created via generative artificial intelligence (GenAI) could be utilized in teaching and learning and to do so ethically. 

They examined the results when users crafted and submitted text-based prompts (called prompt engineering) into the DALL-E 2 platform, which then created images, Pavlik said. The two prompts they used were “Impressionist painting of a purple flower arrangement,” and “surrealist painting of a purple flower arrangement.”

“These prompts and associated visualizations illustrate how the DALL-E 2 platform could be used by students to conduct a comparative analysis of artistic imagery using different styles of art as generated by DALL-E 2,” Pavlik said. “Students could critically assess and discuss how well DALL-E 2 conforms to authentic styles of different styles of painting in the art it generates.” 

Some of their findings were surprising, Pavlik said, including “the widely varying capacity of GenAI in creating different types of visualizations. In some arenas (e.g., impressionist art), GenAI can be quite effective in creating visualizations, but in other areas, especially those more abstract or conceptual (surrealist), it struggles.”

Based on their results, Pavlik said they found DALL-E-2 could be employed as an effective teaching tool in the following scenarios. “For a student of painting, using DALL-E-2 prompts could allow them to work out ideas before beginning a draft of a painting. For young artists experimenting with DALL-E-2, prompts may give them an opportunity to explore other styles or subjects while they are still developing their individual creative voice as an artist. In a classroom setting, DALL-E-2 could function as a vehicle for students to explore various combinations of styles and subjects which otherwise may never intersect.”  

Further, Pavlik explained, “The current investigation suggests a means by which educators can craft assignments utilizing GAI to actively engage students in art creation experiences, even students who do not yet have skills in painting or other techniques endemic to various styles of visual art. Moreover, such GAI-based art-creation experiences offer potentially powerful and ground-breaking opportunities to increase accessibility for students with disabilities to participate actively in the creation of art.”

While the findings of this study suggest that DALL-E 2 can enable students with little or no technical artistic or painting skill to participate in the creation of visual art or art-like materials, Pavlik cautioned that “to participate in this GAI-based art experience, some background in art and art history is essential. To engage in the critical analysis of GAI-created art-like visualizations, students also need relevant knowledge of art and art history that enables the meaningful analysis of the art or art simulations via GAI.”

“The current investigation suggests a means by which educators can craft assignments utilizing GAI to actively engage students in art creation experiences, even students who do not yet have skills in painting or other techniques endemic to various styles of visual art. Moreover, such GAI-based art-creation experiences offer potentially powerful and ground-breaking opportunities to increase accessibility for students with disabilities to participate actively in the creation of art.”

In addition to the study’s findings for art educators, Pavlik said their findings also have significant implications for journalism and media students, and he plans to share this research with SC&I students in both his undergraduate and graduate courses as a recommended reading and as a topic for possible class discussion. 

“As visual representations are a vital part of the journalism and media landscape,” Pavlik said, “it is important that our students learn about the role of GenAI in shaping that visual landscape. Although this study was framed in an art context (e.g., impressionist or surrealist schools of art), the principles of helping students learn to be critical consumers and creators of visualizations in journalism and media are just as essential.”

As students of all ages are most likely going to be interested in GenAI, Pavlik said, “It is important that students have improved literacy in the pros and cons of GenAI. The findings of this research can help in that regard by providing guidance to both students and educators interested in better understanding GenAI and its capacity. This is especially so for students interested in reporting about the arts or in arts-related media or cultural institutions.” 

Learn more about the Journalism and Media Studies Department and major at the Rutgers School of Communication and Information on the website.

Image credit: Pavlik, J., & Pavlik, O. (2024) Art education and generative ai: An exploratory study in constructivist learning and visualization automation for the classroom. Creative Education, 15, 601-61. https://www.scirp.org/journal/paperinformation?paperid=132790