AI Lingo Without the Jargon: A Teacher’s Guide

If you are a teacher in Northern Ireland and have never heard the terms prompt engineering, chatbot, AI agent or repair and maintenance, then you have done an excellent job of avoiding the topic of the moment: AI in education!

AI seems to have developed a language almost as quickly as the technology itself. These terms can make fairly straightforward ideas sound far more complicated than they are.

Teachers do not need to become computer scientists, but understanding some of the basic vocabulary makes it easier to follow the conversation, ask informed questions and decide whether a tool might actually be useful.

We are all at different stages in our own journey with AI in the classroom. Some of us are still finding our feet, while others are more experienced users. Competence builds confidence, so here is my teacher-friendly guide to some basic AI lingo.

Artificial intelligence and generative AI

Artificial intelligence, or AI, is the broad term for technology designed to perform tasks such as recognising patterns or interpreting language.

Generative AI creates new content, including text, images, audio, presentations and quizzes, in response to an instruction. ChatGPT, Microsoft Copilot and Google Gemini are familiar examples.

Model, LLM and chatbot

An AI model is the underlying system trained to recognise patterns and produce responses.

A large language model, or LLM, is trained on very large amounts of text and uses patterns in language to generate a response. It does not understand a subject in the same way a teacher does, nor does it simply retrieve guaranteed correct answers. This is why it can sometimes sound convincing while being inaccurate.

A chatbot is the interface through which we communicate with the model.

Put simply, the LLM is the engine underneath. The chatbot is where we give instructions and receive the response.

Prompt and prompt engineering

A prompt is the instruction, question or information you give to an AI tool.

“Create a quiz about protein” is a prompt. The response will usually be more useful if you also explain the pupils’ age and ability, the course, the format you want and the assessment focus.

Prompt engineering means designing and refining that instruction to produce a more useful result.

Teachers already do something similar when we give clear instructions or success criteria. You can also ask the chatbot to improve a basic prompt for you, then check that the revised version actually reflects what you need.

Output and hallucination

The content AI produces is called the output.

An AI hallucination occurs when a tool generates information that sounds plausible but is inaccurate or invented. It might create a reference that does not exist or include material that is not on a specification.

A polished output should never be confused with a checked, classroom-ready resource.

Tokens and context window

AI models process information in units called tokens, which may be whole words, parts of words or punctuation.

The context window is the amount of information the model can consider at one time. You can think of it as the working space available for the task.

In a very long conversation, earlier details may eventually fall outside that space.

Multimodal AI

Multimodal AI can work with more than one type of information, such as text, images, files or audio.

For a teacher, this might mean uploading a worksheet for review, asking questions about a diagram or turning a set of notes into a presentation.

AI agent

An AI agent can be given a broader goal and, depending on the tool, may be able to plan several steps and use available tools to help complete a task.

For example, it might search for information, organise what it finds and produce a summary.

It still needs clear boundaries, reliable sources and human oversight.

Turning off training

When someone advises you to turn off training, they generally mean changing a privacy setting so that new conversations are not used to help improve the company’s AI models.

In a personal ChatGPT account, this setting is currently called Improve the model for everyone. On the web, go to Profile → Settings → Data Controls. On the mobile app, open the side menu, tap your profile and select Data Controls. Switching it off does not remove your chat history.

[CHECK: confirm the current ChatGPT setting name and menu path before publishing, as product settings can change.]

This is useful, but it does not make it appropriate to upload identifiable pupil information, confidential documents or sensitive school data. Teachers must still follow school data-protection procedures and current C2k guidance.

[CHECK: confirm the latest C2k guidance and whether you want to link directly to it.]

Settings may also differ in organisation-managed AI tools.

Repair and maintenance

Repair and maintenance is how I describe the work that often follows AI’s first response.

Repair means correcting inaccuracies, omissions, unsuitable language or poor design. It might involve aligning a resource with a CCEA specification, adapting it for a particular class or removing misleading material.

Maintenance means continuing to review and update the prompt or resource. Specifications, guidance and AI tools change, and something that worked with one class may need adapting for another.

AI can reduce workload, but its first response is not always oven-ready. Repair might take a few minutes or, occasionally, almost as long as creating the resource from scratch.

The teacher still has to decide whether the final result is accurate, appropriate and genuinely useful.

You do not need to know everything

The aim is not to memorise every technical term, but to understand enough to join the conversation without feeling that everyone else has been given a secret dictionary.

Then we can concentrate on what actually matters:

·       What teaching problem am I trying to solve?

·       Will this improve learning?

·       What needs to be checked, repaired or adapted?

With technology moving so quickly, even experienced AI users have to work to keep up with the latest terminology.

What AI terms or concepts have you come across that you would like broken down? Share them in the comments.