Section 4 – Practice
In this section, your project is to design a small AI-assisted solution to a real problem, build a simple prototype, and plan how you would test and improve it.
Begin by choosing a real problem you care about. It could be helping younger kids practice reading, organizing homework and deadlines, encouraging better recycling at home or school, or reminding people to take short breaks during long study sessions. On paper, write a short description of the problem and clearly say who your main user is. Then create a six-box storyboard: in the first box show the user struggling with the problem, in the next few boxes show what they try, where AI helps, and how things change, and in the last box show what success looks like.
Write down one or two specific things you want AI to do in your idea, such as generating hints, checking answers, recognizing objects, or suggesting next steps. Put these in a short list called “AI’s Job” next to your storyboard.
If you have access to Scratch, you can turn your storyboard into a basic interactive prototype. Go to:
https://scratch.mit.edu/
Create sprites to represent your user, the AI helper, and any key objects. Use code blocks to move through your idea’s steps, ask the user questions, and show responses that pretend to be AI behavior. For example, a reading helper might ask the user to type a word and then show a simpler explanation from a list, simulating how an AI explanation tool would work.
If you have access to Teachable Machine, you can add a small real AI element to your prototype. Visit:
https://teachablemachine.withgoogle.com/
Train a simple image or sound model related to your project, like recognizing two types of recyclable items, or detecting a clap to start a break. Many tutorials show how to export a model into web or Scratch-like environments; if that’s available to you, connect your model so your project responds to real webcam or microphone input.
If you have access to a text-based AI tool or a Python environment, you can work on your prompts and testing. On a safe chat interface, try three different versions of a prompt that your project might use, such as “Explain this word in a simple way for a 9-year-old.” Compare the outputs and decide which prompt gives the best results. In Python (for example on Replit: https://replit.com/), you can store your best prompts in a file and write a very short script that prints them or simulates how your tool might call them. Finish by writing a short test plan describing two normal ways you expect people to use your project and two unusual or “edge-case” ways—and three questions you would ask real users to guide your next improvements.