AI For Education And Teaching

A useful starting point is to treat AI as a capable assistant rather than an expert colleague. It can suggest examples, reorganise information, adapt the reading level of a passage, or offer several activity ideas. It does not understand a class in the same way a teacher does. It has no direct experience of a learner’s confidence, home situation, prior knowledge, or non-verbal response in a lesson.

Ask yourself: if an AI-generated explanation were placed on a worksheet, what would need checking first? The answer might include factual accuracy, reading level, cultural suitability, accessibility, and alignment with the learning goal. The practical takeaway is simple: AI may speed up preparation, but the educator remains responsible for the result.

What AI Can and Cannot Do

Artificial intelligence (AI) is software designed to perform tasks that usually require human intelligence, such as recognising patterns, generating language, making predictions, or classifying information. In education, common tools include chatbots that generate text, speech-to-text systems, image generators, automated feedback tools, and platforms that recommend resources.

Generative AI creates new content from a user’s instruction, often called a prompt. It may produce a lesson outline, discussion questions, a model paragraph, or a simplified explanation. However, “new” does not mean independently understood or guaranteed to be true. A language model predicts likely sequences of words from patterns in its training and its current instructions. It can produce a convincing answer without checking reality.

Hallucination is an incorrect or invented output presented as though it were accurate. For example, a teacher might ask for five sources on a historical event and receive titles, authors, or website links that do not exist. The wording may look academic, but the references still need to be opened and checked. If the tool provides a date, quotation, statistic, or research finding, treat it as a claim to verify.

AI is often useful for breadth. It can produce ten possible starter activities when a teacher has only thought of two. It can rewrite an explanation in simpler language, create a first draft of a rubric, or generate alternative examples for learners who need more practice. It is less reliable when the task depends on precise facts, subtle judgment, local context, or knowledge of an individual learner.

Consider an AI-generated explanation of fractions. The tool may correctly state that 3/4 is greater than 2/3, but it might choose an example involving food that conflicts with dietary needs or cultural expectations. It might also use visual language that does not work for a learner with a visual impairment. The mathematical statement can be correct while the teaching choice is poor.

Bias is a repeated pattern of unfairness or imbalance in data, design, or output. An AI system may describe some occupations using masculine language, assume that a “standard” family has two parents, or provide examples that centre one culture while ignoring others. Bias is not always obvious. Ask whether the examples represent the learners, whether any group is described as a problem, and whether the task gives all learners a fair chance to succeed.

Automation bias is the tendency to trust a computer-generated answer simply because a system produced it. A teacher may accept an AI-generated behaviour comment because it sounds professional, or a learner may believe an automated score is more objective than teacher feedback. Professional language is not evidence of correctness. A review process is still needed.

Human-in-the-loop means that a person remains involved in checking, deciding, and taking responsibility. In teaching, this might mean reviewing every generated question, testing an activity with the intended age group, and editing feedback before sharing it. Human review is not a final formality. It is part of the teaching process.

Privacy requires particular care. Do not paste identifiable learner information into a public AI tool unless your organisation has approved that use and the necessary protections are in place. A name, email address, student number, medical detail, or distinctive personal story may identify someone. Replace details with general labels, such as “a 14-year-old learner who finds extended writing difficult,” and check your school, college, or training provider’s policy.

A practical check has four questions:

1. What is the learning purpose? 2. What could be wrong or harmful in this output? 3. What evidence will I use to check it? 4. What decision must remain mine as the educator?

Suppose AI creates feedback on a learner’s paragraph. The teacher should check whether the feedback matches the stated success criteria, whether it identifies a real issue, whether the tone is respectful, and whether it gives the learner a useful next step. The teacher should not assume that a confident paragraph score reflects the learner’s actual understanding.

The key distinction is between assistance and authority. AI can offer possibilities; it cannot take responsibility for accuracy, fairness, safeguarding, or the relationship between teacher and learner. Before using an output, ask: “Would I be comfortable explaining and defending this decision to the learner, a colleague, or a parent?” If not, the output needs more checking or should not be used.

A Worked Example: Checking an AI-Generated Lesson Starter

A teacher asks an AI tool to create a ten-minute lesson starter for a Year 8 science class on separating mixtures. The learning goal is: “Learners can choose a suitable method for separating a mixture and explain why it works.” The tool returns this activity:

> “Show learners a mixture of salt, sand, and iron filings. Ask them to use a magnet to remove the salt, filter the remaining mixture, and evaporate the water to collect the sand.”

The activity sounds plausible, but it contains several problems. The teacher needs to check the science, the equipment, the safety, and the match with the learning goal.

1. State the intended outcome. The target is not simply naming methods. Learners must connect a property of a substance to a separation method. The important properties are magnetic attraction, solubility, particle size, and boiling point.

2. Check each proposed step. A magnet can remove iron filings, but not salt. If water is added, salt dissolves while sand does not. Filtration can then separate the sand from the salt solution. Evaporation can recover the salt, not the sand. The AI output has reversed the final result.

3. Correct the sequence. The teacher revises the process: - Use a magnet to remove the iron filings. - Add water to dissolve the salt. - Filter the mixture to collect the sand. - Evaporate the water to recover the salt.

4. Check the numbers and resources. The teacher has 24 learners, six working groups, and six filter funnels. Each group can receive 5 grams of sand, 5 grams of salt, and a small amount of iron filings. The activity is planned for ten minutes, so evaporation cannot realistically be completed during the starter. The teacher decides to demonstrate the evaporation stage or show a prepared sample.

5. Check safety and accessibility. Learners need eye protection, careful handling of iron filings, and clear instructions about hot equipment if evaporation is demonstrated. A learner with limited hand mobility may need a partner or adapted equipment. The teacher prepares a labelled diagram so that the sequence is not carried only through spoken instructions.

6. Check the questions. Instead of asking only, “Which method comes next?” the teacher asks: - “Why does the magnet remove the iron filings?” - “Why does the sand remain on the filter paper?” - “Where has the salt gone before evaporation?” These questions reveal whether learners understand the properties involved.

7. Decide what to use. The AI output is not discarded completely. It provided a useful mixture and a starting structure, but the teacher corrected the science, adjusted the timing, considered safety, and redesigned the questions.

Final result: the teacher uses a corrected four-stage separation sequence, demonstrates evaporation rather than promising learners will complete it in ten minutes, and checks understanding through property-based questions.

This example shows why AI review must be active. The problem was not a strange sentence or an obvious error. It was a small scientific mistake embedded in an otherwise usable activity. A teacher who copied the plan might teach the wrong concept and confuse learners. A teacher who checks the reasoning can still benefit from the tool’s speed.

The practical takeaway is to review AI output at three levels: truth, teaching, and context. Is it accurate? Does it help learners reach the objective? Does it fit this group, this equipment, this time, and this setting?

Guided Practice and Answer Key

1. An AI tool creates a reading passage that includes three statistics but gives no sources. What should you do before using it? Hint: Treat every statistic as a claim. Look for trustworthy original sources and check whether the figures match the intended age group and topic.

2. A chatbot suggests giving individualised feedback after you paste eight learners’ names, diagnoses, and writing samples into a free public tool. What risks should you identify? Hint: Consider privacy, consent, organisational policy, and whether the information could identify or disadvantage a learner.

3. AI produces ten discussion questions about a novel, but all ten focus on the main character’s choices. How might this limit learning? Hint: Check whether the questions cover the learning objective, supporting characters, language, structure, evidence, and different levels of thinking.

4. A generated rubric gives a learner a low score because the response uses short sentences, although the objective is to explain a scientific process accurately. What should the teacher check? Hint: Compare the rubric with the stated success criteria. A feature that is not part of the objective should not quietly control the judgment.

5. Write one prompt for an AI tool that asks for activity ideas while keeping professional judgment with the teacher. Hint: Include the age group, learning objective, time, resources, and a request to state assumptions and possible risks.

Answer Key

For Question 1, verify the statistics using reliable sources before publication or teaching. For Question 2, remove identifying information and follow the organisation’s approved privacy process; do not assume a public tool is suitable. For Question 3, broaden the questions so they serve the objective and represent more than one viewpoint. For Question 4, revise the rubric so it measures accurate scientific explanation rather than an unrelated writing preference. For Question 5, a strong prompt gives clear teaching context and asks for options, limitations, and checks rather than requesting one unquestioned answer.

AI literacy is not the ability to accept every new tool or reject it altogether. It is the habit of asking what the system has done, what it may have missed, and what the educator must decide. Used in that way, AI can reduce routine preparation while keeping accuracy, fairness, privacy, and learner needs at the centre of teaching.

End of chapter one. 7 more chapters in the full book.