AI Teaching Strategy 1: Art Vocabulary
Building Vocabulary through Practice
While many teachers fight new battles with AI over age-old issues with maintaining Academic Honesty, we are looking for creative ways teachers can use this tool to improve instruction. Generative AI offers opportunities to test the limits of students' developing vocabulary with visible results. In addition, learning to ask questions of AI provides new avenues for deepening students' observation and description skills. Here, we'll look at some excellent ideas that Konstantina Kapanidou has been workshopping in her classroom in Greece.
Strategy 1: Descriptive Writing and Subject-Specific Vocabulary
Eye rolls and sighs are routinely pervasive when students are asked to write about art. But what if we used AI to gamify this rather "boring" task? What better than to employ our new friend AI?! Ideogram is a free generative AI tool that allows users to create images based on descriptors and parameters (quality, ratio, style, color palette, etc.). The free plan (as of 2024) includes a default setting of five searches, with each result yielding four newly generated images (20 images in total).
Starting Points
A) After students have completed a new artwork, to stretch their vocabulary development and descriptive powers, ask them to open up Ideogram to "generate" an image that matches their recently completed artwork.
OR
B) Without using names or titles, ask students to describe an artist's artwork they're studying as part of their comparative study or in preparation for a new artwork. How close can they get to generating something that looks like the original?
Iterate
After students have generated their first batch of images, they should "iterate" or refine their search. They will likely need to add additional details, including the position or placement of the subject, cropping, foreground/background, perspective, color, value, textures, etc. With Ideogram in the free version, students only get five chances to improve their search. They won't be able to achieve a final result that looks exactly like their actual artwork or another artist's; however, the images produced should progressively get closer to the visuals they observe in the artwork they're describing.
Regardless of which option students use as their starting point, as they iterate, they should document 1) the image results produced by the AI tool and 2) the corresponding text prompts they feed to AI.
Vocabulary Reflection
Using their documentation, students can analyse their writing and reflect upon how they improved. Here are a few guiding questions to share with students:
- What descriptors was I lacking in the first iteration that improved the result in round 2?
- What visual arts terminology did I add to my descriptions to improve my results?
- What can I apply to future art observation/analysis activities based on my findings from today?
You can share the companion student facing page AI Activity 1: Art Vocabulary directly with your students in student access.
Here is an example of Strategy 1, based on one of my (Shannon) teaching examples using Procreate for a digital arts class. If the image looks familiar, I produced this artwork after generating ideas by applying the SCAMPER strategy, combining artworks by Gary Hume and Georgia O'Keeffe. I ran my image through Ideogram five times. The image progression below shows the "closest fit" from each iteration. As you can see, compositionally, the generated images grow similarly to the original artwork seen on the left from the first trial to the last without ever becoming a perfect match.
Example AI Iterations:
Here are examples of the text I fed into Ideogram for Tests 1 and 5 and the resulting four images provided for these two tests. I could not continue beyond the 5th iteration as I only have a free version; however, the limited number of tests I could attempt enhanced my motivation to sharpen my description at each stage.
Test 1: "A skull in the foreground inspired by Georgia O'Keeffe's painting "Red, White & Blue" with wine-red roses in the middle-ground inspired by Gary Hume's artwork "Two Roses." The background should be a light blue gradient."
Below, I've shared the starting point of a much more elaborate prompt (quite lengthy!) and highlighted examples of my subject-specific terminology improvements that were necessary additions to reach the closest match.
Test 5: "A cow's skull, seen from a frontal position, is in the foreground on the right side of a composition, painted in off-white, ivory tints with some slight shadows. Only one half of the skull is visible due to cropping. The skull is smoothly shaded without a lot of texture or detail. The horn is also a creamy white. The style is inspired by Georgia O'Keeffe's painting "Red, White & Blue." Red roses with cadmium, vermillion and alizarin crimson hues are seen close up in the middle-ground. There are four roses. The roses are only seen from a frontal view in an illustrated style with loose brush strokes. There is one rose towards the center of the composition, but it is slightly covered up by the skull on the right side..."
Interestingly, this AI tool didn't fully grasp some aspects of my description (i.e., the light blue background with purple splatters). This kind of discrepancy offers further fodder for discussions with students about the limitations of AI and the variables of how certain aspects of descriptions may be (re) interpreted, reprioritized, and devalued by the tools.
Further Resources
At the time of writing this article, here are some free resources that are breaking new ground for students wishing to generate images based on prompts and pre-existing imagery:
- https://openart.ai/
- ideogram.ai/
- https://www.bing.com/images/create
Not all AI-image generators offer the same quality or accuracy, so we recommend you play around a bit with the possibilities before selecting which tool you think will best fit your classroom and strategy.
References
“A skull in the foreground inspired by Georgia O'Keeffe's painting 'Red, White & Blue' with wine-red roses in the middle-ground inspired by Gary Hume's artwork 'Two Roses.' The background should be a light blue gradient."” prompt. Ideogram, 2.0 version, 1. Dec. 2024, https://ideogram.ai/.