What good prompting looks like for teachers

As a teacher, you are well aware of what happens in the classroom when your instructions fall a bit short. You see the look on the faces of your students. Then they look at each other to see if they’re the only one who is confused.

It is still too early in the development of AI tools for the technology to give you a look after a confusing set of instructions. Perhaps one day soon. For now, what you are going to get is an output that doesn’t match your expectations and takes more time to refine.

Giving good classroom instructions is an art, and one that needs to be practiced. The same is true for AI. So let’s discuss how we can move from beginner to genius level prompters.

Prompting Basics

Just like language learning starts with simple vocabulary and grammar, prompting also has its fundamentals. Here are some better-than-nothing techniques that will make an immediate difference.

Give the AI a role. Start with something like “You are an experienced materials writer for advanced level ELT materials, specifically for university level TESOL programs.” This helps the model to narrow down the text which it draws from to generate the output. Without a role, it will have to guess the register, style, and depth of knowledge. It’s the difference between asking a convention hall full of teachers a question versus approaching the one teacher you know who specializes in that area.

Add your class context. Include all the information you collected from the student needs analysis you conducted at the beginning of term, and anything else you gleaned from your students since then. Some obvious information not to miss is level, L1, student goals and preferences. Also include administrative information like defined learning outcomes and timelines. The more the better.

Tell the AI what TO do, not what not to do. Instead of “Don’t use complex vocabulary or long sentences,” try “Use A2 level vocabulary and limit sentence structure to simple and compound sentences.” Telling the AI to NOT do something is like telling it to do anything else but that. Don’t leave your requirements undefined.

Next Level Prompting

Once you are comfortable with the basics, it’s time to step up your game. Let’s get more intentional and repeatable.

Utilize multishot prompting. This is a technical way to say sharing more than one example of what you want from the AI. If you are wanting a lesson plan, then share a few examples of high quality lesson plans that you have written. A few examples of exactly what you’re expecting is more useful than any description.

Practice prompt chaining. AI is capable of a lot, but you can’t expect it all at once. Asking it to write your syllabus, all lesson plans and materials, as well as the final exam, all at once, is not going to yield what you hope. Start with one output. When that is finished, move to the next. This will ensure you have the full attention of the AI for each task.

Explain the why. Instead of “keep sentences short,” try “keep sentences short because my B1 level students struggle with embedded clauses.” In doing so, the AI will be able to extrapolate from this constraint to other areas, not just the one you named.

Prompt Like a Pro

Because we are educators who believe in reaching that next level, here are some more advanced techniques that you can use to impress the other teachers in the breakroom.

Include XML tags to separate context, task, and constraints when including each in the same prompt. This will help the AI to parse complex prompts clearly. Here’s an example:

<context>I teach a B1 adult general English class. Students are mostly French speakers.</context> <task>Write a 200-word reading text on the topic of environmental sustainability.</task> <constraints>Use present simple and present perfect only. Avoid idioms.</constraints>

Save a master prompt. Once you start getting the results you want, save that prompt in a document. Identify the placeholders within the prompt, e.g. topic, level, language point. Then, when it is time to generate a new lesson plan, copy and paste in your template. All you have to do is fill in the lesson specific information.

Ask the AI to evaluate your prompt. The AI will look for gaps or ambiguities. Then you can make those improvements before it generates the output. Iterate as you go to continue to improve the output.

A final thought

Imagine the AI model is a brilliant new teacher on the team. They’re not just brilliant. They are new. They are not just new. They are brilliant. Can they do an amazing job? Absolutely. Do they need to fully understand the context, constraints, and customs in order to be amazing? Also. Remember, the AI can only predict the best response based on what you put in the prompt. The richer the context you provide, the more accurate the output. Be patient, supportive, and before you know it you will have a top notch collaborator.