The Missing AI Literacy Conversation in Schools

I am seeing a lot of conversations right now around AI tools, AI grading, and whether teachers and students should be using AI.

I think there is an AI literacy piece missing from a lot of these conversations.

As I have said before, I am not a big fan of the term “AI.”

It makes the tool sound like a magic box. Depending on who is talking, AI is either put on a pedestal or demonized completely.

I do not think either view is helpful.

AI is not magic.

It is also not something schools can simply ignore.

At its core, generative AI is a sophisticated machine learning system that uses patterns from training data, user input, and available context to generate a likely response.

That can be very powerful.

It can also be wrong.

So the real question is not:

Should teachers or students use AI?

The better question is:

Where does AI actually fit into the workflow?

That distinction matters.

---

“AI Tool” Is Too Broad

One of the problems with the phrase “AI tool” is that it is too general.

Schools have used AI-related tools for years.

Spam detection is one example.

Spell check is another.

Search engines, grammar tools, recommendation systems, and plagiarism checkers all use different kinds of algorithms and pattern matching.

But when most people talk about AI right now, they are usually talking about generative AI.

That is different.

Generative AI can produce new text, images, summaries, questions, outlines, feedback, and other outputs based on the information it is given.

Instead of just correcting a misspelled word or helping find a webpage, it can help rewrite an entire email, summarize a document, create a draft lesson, or generate practice questions.

That is useful.

But it does not remove the human part of the process.

And when we try to remove the human part, that is where many of the problems start.

---

AI Works Best Inside a Workflow

The power of AI is not that it can do everything.

The power is that it can support specific parts of a workflow.

That is how we should think about it in schools.

Not:

What can this tool do?

But:

What task can this tool responsibly support?

That is a much better question.

A tool should either:

• replace a repetitive task
• help organize information
• give faster feedback
• act as a second set of eyes
• help identify patterns
• or support a task the teacher is already doing

If it does not improve the workflow, it may just be another disconnected tool teachers have to manage.

That is not helpful.

Teachers do not need more disconnected tools.

They need better systems that actually fit the school day.

---

AI Grading Needs to Be Treated Carefully

AI grading is a good example.

Is there a place for AI in the grading workflow?

Yes.

But it depends on what we mean by grading.

If we mean AI should become the final judge of student learning, I think that is a problem.

A model should not be treated as the source of truth.

It is looking for patterns. Those patterns may be useful, but they are still based on training data, the prompt, the rubric, the student response, and the context it has been given.

That means the output can be helpful.

It can also be wrong.

So instead of asking whether AI should grade, I think we should ask:

What part of the grading process can AI safely support?

That might include:

• identifying possible misconceptions
• checking whether a response addresses parts of a rubric
• organizing student responses by pattern
• helping draft feedback
• generating follow-up questions
• or flagging work that needs closer teacher review

That is different from handing the grade over to the model.

AI can support the grading workflow.

But the teacher still needs to own the judgment.

---

Example 1: Checking for Understanding

Let’s say I want to know whether students understand how energy transfers from the Sun to plants and then to consumers.

There are a few ways AI could support that workflow.

One option would be to give students open-response questions and use an AI model to help compare their answers to a rubric or expected response.

This could help identify whether the student included key ideas, missed an important concept, or showed a possible misconception.

Some systems may use embeddings or other similarity methods to compare meaning between the student answer and expected concepts.

That can be useful.

But it still needs teacher review.

The model can help identify patterns.

The teacher still decides whether the student showed understanding.

That is the difference.

---

Example 2: Using AI to Support Deeper Assessment

Another option is to use AI after students complete a larger assignment.

For example, if students complete a research project, an AI tool could help identify the main claims, vocabulary, evidence, and ideas presented in the paper.

Then it could help generate a follow-up assessment based on that specific student work.

I call these AEQs:

Application and Explanation Quizzes

The goal is not just to see whether the paper met the rubric.

The goal is to see whether the student can explain and apply the ideas they submitted.

That changes the assessment.

Instead of only asking:

Did the student turn in a polished product?

We can ask:

Can the student explain, apply, and defend what they submitted?

That gets much closer to actual learning.

This is where AI can be powerful when used correctly.

It can help create a second learning moment.

But again, the AI is supporting the workflow.

It is not replacing the teacher.

---

Teacher Use and Student Use Are Not the Same

When it comes to teachers and students using AI, I do not think the question should be whether they should use it or not.

The better question is:

How should AI be used in the workflow?

Teachers might use AI to:

• give feedback faster
• adapt a lesson
• organize assessment data
• identify student patterns
• draft communication
• create first versions of materials
• or help prioritize instruction

But teachers should not use AI as the final source of truth.

They should not automate the parts of the job that require professional judgment.

Students also need to learn how to use AI responsibly.

They can use it to:

• gather starting points for research
• see alternative explanations
• ask questions about a topic
• organize their thinking
• challenge an answer
• or check whether their explanation is clear

But they also need to learn how to recognize when AI is wrong.

They need to verify sources.

They need to understand the material before using AI to speed up the process.

Just like students should learn how to find and evaluate sources before relying on a tool to help gather them, they should also learn how to think before using AI to speed up their thinking.

AI should enhance learning.

It should not replace the work required to earn understanding.

---

Human Over the Loop With Governance

The key to all of this is governance.

Teachers and students need to use AI in ways that keep humans responsible for the process.

I think about this as:

Human Over the Loop with Governance

That means the human is not just checking the final output after the AI is done.

The human is setting the rules for how the tool is used in the first place.

That includes asking:

• What task is AI supporting?
• What information is it allowed to use?
• What output is expected?
• What must be checked by a human?
• What should never be fully automated?
• Where does teacher judgment matter most?

That matters because AI should only be given autonomy when the process can be governed, checked, and corrected by humans.

The more a task directly affects a student, the more human review matters.

---

Final Thought

AI is not the problem by itself.

The problem is using AI without understanding what it is doing or where it fits.

If we treat AI like a magic answer machine, we will create problems.

If we treat it like a tool inside a clear workflow, it can help.

That is the AI literacy piece I think schools need more of.

Not just:

What can AI do?

But:

What should AI do, where does it fit, and how do we keep humans responsible for the decisions that matter?