Innovation and Imagination · Lesson 2
Students write and revise prompts in the FRACTAL AI Art tool, compare how descriptive language changes what the AI generates, and learn how training data can lead to errors and bias.
45 minutes 
Open lesson slides
Students continue to explore how AI tools generate images by reproducing patterns from images and text created by humans. Students generate images in FRACTAL’s Digital Studio, explain how they improved their prompt, and note whether they observed any signs of bias in the AI’s output. Students see examples from different AI image generators and analyze the differences in a discussion about AI training.
Students then reflect on how they can adapt or change one of their AI-generated images to better express their own ideas and create artwork inspired by the image.
Full Expedition standards, by grade and lesson
5–7 minutes · Slides 2–4
15 minutes · Slides 5–18
25 minutes · Slides 19
5 minutes · Slides 20–21
The lesson deck with its speaker notes — what to say and what to do on each slide, with suggested timing. Slide numbers match the deck in Google Slides.
1 min
Say
This slide includes quick links to today’s lesson materials.
Do
Teacher-facing slide. Before class, open the Lesson 2 materials and Fractal Studio AI Art Tool. Make sure student devices are ready and students have access to their login information. This slide can be skipped when presenting to students.
1 min
Say
Today we’re continuing Expedition 3: Imagination and Innovation. Last time, we explored how people and AI can both create images from words, but in different ways. Today, you’re going to have a chance to work directly with an AI image-generation tool and see how the words you choose affect what it creates.
Do
Teacher-facing title/reference slide. This slide may be skipped when presenting to students.
5 min
Say
Have you ever used an AI image generator before? If you have, think about what you used it for and what the experience was like. If you haven’t, think about why you haven’t tried one yet. Now consider this second question: How do you feel about making images using AI? You might be curious, excited, unsure, skeptical, or something completely different.Take a minute to write or sketch your response. Then turn to someone near you and compare your experiences and reactions.
Do
Give students 1–2 minutes for independent reflection, followed by partner discussion.Invite 2–3 students to share different perspectives. Normalize a range of experiences and attitudes; students do not need prior experience with AI tools to participate successfully.
1 min
Say
We’ll start by thinking about our experiences with AI image generators. Then we’ll learn more about image generation and prompts, explore how descriptive words affect the results, and finish by creating and testing our own prompts in Fractal Studio.
2 min
Say
Image generation is a computer process that creates image content.Some AI tools are programmed to generate images from words. Instead of searching for an existing image, the tool creates an output based on the words you provide and the patterns it learned during training. Last time, we saw that AI does not imagine in exactly the same way people do. It uses patterns and relationships it has learned between words and images.
2 min
Say
The words we give an AI tool are called a prompt. A prompt is text that tells the AI program what kind of response or image we want. Think back to our drawing activity last time. When someone told you to draw a monster chasing a dog, those words were essentially your prompt. You interpreted the words and decided what to draw. Today, we’ll see what happens when we give similar kinds of instructions to an AI tool.
2 min
Say
Before looking closely at the result, read this prompt: Primary color, retro style, Tetris-inspired cat jumping over a green pipe. Make it low resolution.What information did the prompt give the AI? We have a subject—a cat. We have an action—jumping. We have colors, a style, another object, and information about resolution. Now look at the result. Did it match what you pictured? Why might the cat be shown jumping sideways instead of toward the viewer?
Do
Invite several quick observations. Guide students toward the idea that a prompt gives the tool information but does not specify every detail. If a relationship, position, viewpoint, or other feature matters, the prompt may need to describe it more clearly. The prompt is predicting what we want to see based off patterns it has been taught to pay attention to.
3 min
Say
Look at these three prompts. They all start with a cat, but each one adds different information about how the image should look. Cat in a Kandinsky style. Geometric abstract cat. Cat with shapes for a body and no face. Notice how changing the descriptive language changes the output. If you wanted an image that looked like anime, how would you describe the visual features of that style without simply saying anime? What kinds of colors, lines, shapes, expressions, or details might you include? Knowing the language of art helps us critique it.
Do
Invite students to generate descriptive words. Encourage language about visible features such as: bold or thin outlines large expressive eyes exaggerated facial expressions bright or limited colors simplified or detailed backgrounds
1 min
Say
Being descriptive can also make the process more efficient. The clearer you are about what you want, the more likely you are to get a useful result without generating image after image.That can save time and reduce unnecessary use of the computers and resources required to generate images. The goal isn’t necessarily to make the longest possible prompt. It’s to include the details that matter.
1 minutes
Say
Here are two images that were generated using the FRACTAL AI Art Tool. The prompts used for each were: “A monster chasing a dog”, and “A happy monster chasing a dog”. What are your first impressions? What do you think about the colors? The shapes? Why is the image from the last class darker and spookier?
Do
Invite 2–3 observations. Look for differences in: brightness facial expressions body positions colors background overall mood Reinforce that descriptive words can influence many parts of an image, not just one object.
2 min
Say
Now compare these two images generated from similar ideas. Focus on three parts: the monster, the chasing action, and the dog. How are those parts represented differently? Why might one output look more realistic while another looks more cartoon-like or simplified?Both tools are making use of patterns they learned during training, but different AI tools may have learned from different data or may process the same prompt differently.
Do
Ask students to name 2–3 descriptive words for each image. Emphasize that the same prompt does not guarantee the same output across different AI tools—or even across multiple generations from the same tool
1 min
Say
The information used to train an AI tool matters because it influences the patterns the tool learns. If the training examples are limited, incomplete, or uneven, the results can also be limited or inaccurate. This is one way bias can appear in an AI system. Look closely at this dog. It is not completely accurate, but we can still recognize features associated with a dog. When we use AI, we should not assume that every image it generates is accurate simply because it looks convincing. The ways a model are trained can be thought of as bias. The FRACTAL tool was trained only on open-source images that living artists had the right to withdraw their work. Given the color and and style, we might wonder whether the model was trained on cartoons or images with bright colors?
Do
Keep the focus at a high level: the data and patterns available to a system can influence the outputs it produces.
2 min
Say
AI tools can sometimes connect patterns in surprising or inaccurate ways. Here’s a famous visual example: chihuahuas and blueberry muffins can share features such as round shapes, dark spots, and similar colors. An AI system working with visual patterns can sometimes confuse things that share features. When an AI-generated image looks wrong or strange, that isn’t the tool being silly on purpose. It is a reminder that the system is working with learned patterns and predictions—and those predictions are not always correct.
Do
Ask students to identify one visual similarity that could make the two categories easy to confuse. Use this slide as a transition from understanding AI outputs to experimenting with the tool themselves.
1 min
Say
Now we’re going to take a quick tour of the FRACTAL AI Art Tool before you experiment with it yourselves. Watch first so you know where the main features are. Then you’ll have time to explore.
Do
Have students log in to FRACTAL Studio, but ask them to pause while you demonstrate.
1 min
Say
This is where you can turn text into images using AI. If you type a prompt in, you will get an image interpretation of your prompt.
Say
If you’re not feeling very inspired, here are some suggestions for things you can generate.
1 min
Say
On the right are style choices. A style can give the tool additional information about the visual qualities you want in the image.You can experiment with different styles and compare how the same basic subject changes when you make a different artistic choice.
Do
Show students where to select a style. If possible, demonstrate how the same simple prompt changes when one style option is changed.
1 min
Say
The center of the tool is the image generator. This is where you turn text into an image. You type your prompt here, select Generate, and the AI tool creates an image based on its interpretation of your words. Remember: the image is an interpretation of the prompt, so it will not likely match the picture you had in your mind.
Do
Demonstrate where to type the prompt and where the generated image appears.
No speaker notes for this slide.
Say
Let’s bring together what we discovered today. A prompt gives an AI tool information about what we want. The words we choose—subject, description, mood, style, position, and other details—can influence the image we receive. But even a detailed prompt does not guarantee a particular image. AI tools generate outputs using patterns learned during training, and they can create results that are unexpected, inaccurate, or different from what we imagined. That means you still have an important role. You decide what to ask for, how to revise your prompt, which results are useful, and what you want to do with the images you generate.
Do
Ask students to share one prompt change that noticeably changed their image.
No speaker notes for this slide.
Slide 1 of 21
Materials & preparation
Students engage with AI image generation through discussion, visual comparison, writing, and hands-on experimentation, providing multiple entry points for students with different levels of prior AI experience.
The prompt-planning categories — Subject, Describe, Emotion, and Style — break prompt construction into smaller parts and provide structure for students who need support getting started. Built-in prompt suggestions and style options in the FRACTAL AI Art tool give students additional support while allowing students who are ready for greater independence to write their own prompts.
The Challenge Cards provide structured practice with different levels of openness. Students may borrow effective descriptive language from a neighbor and revise their own work rather than needing to solve each challenge independently.
