Welcome new subscribers of Real-World AI for Teachers! My work this summer has increased the numbers of educators who have subscribed to this newsletter, and I’m excited to continue providing my curated collection of news stories and classroom-tested lessons to so many teachers. I appreciate you putting your trust in me!

On October 14, attorneys general from eight states filed lawsuits against OpenAI, alleging that ChatGPT harms young users, encourages dangerous behavior, and misleads users about its safety. The suits seek stronger safeguards, age verification, data protections, and other restrictions. OpenAI says it has added youth-focused protections and will continue improving them.

Teachers at the education company Multiverse reported significant stress after the company began using AI to transcribe and evaluate their teaching sessions, flagging things such as response times and speech patterns. The company says the AI supports quality assurance and that humans still make performance decisions, while employees and union representatives have raised concerns about transparency and surveillance.

There is no doubt in my mind that this is the future of teacher evaluation, at least in part. As we try to help our students understand the importance of the “human in the loop” assessing AI-generated information, we must also push back against any high-stakes evaluation of people (students or teachers) solely by machines. Personally, I wouldn’t mind formative feedback from an AI that analyzes video/audio recordings of lessons. But it’s a slippery slope to full-on human-less teacher evaluation.

Google.org announced a $4 million grant to Digital Promise to expand free AI professional development for teachers, school leaders, and college faculty. The training is part of the Google AI Educator Series and is designed to help educators build confidence using AI tools, understand responsible use, and think about how AI fits into teaching and learning.

I like the idea of making AI training more available, especially when so many teachers are still being asked to figure this stuff out on their own time. But there’s also an important distinction here: learning about AI is bigger than learning how to use one company’s AI tools. The best professional development should help teachers make good instructional decisions no matter which logo is on the chatbot.

AI can help students work faster, explain confusing ideas, organize information, brainstorm possibilities, and revise their work.

It can also quietly remove the exact thinking we wanted students to practice.

That makes one skill increasingly important: knowing the difference.

For this activity, students look at several possible ways someone might use AI during an assignment and decide whether each use supports learning or creates a shortcut around it.

The goal isn’t to create a universal list of “good” and “bad” AI uses. It’s to help students build the judgment to ask:

What thinking am I supposed to be doing here, and is AI helping me do it or doing it for me?

Start with an assignment students already understand.

It could be:

Then give students several possible AI uses.

For example:

Possible AI uses:

Students then sort each use into categories such as:

Helps Me Think
Helps Me Work
Does Too Much

The sorting matters less than the reasoning.

For each choice, students should have to explain:

What part of the learning is still mine?

That one question changes the conversation.

Two students may even classify the same AI use differently, and that can be useful.

If a student already understands a concept well, asking AI to polish the wording might simply help with communication. If the learning target is specifically about writing clear explanations, however, that same AI use could bypass the skill being assessed.

Context matters.

That’s exactly the judgment we want students to develop.

The best examples probably shouldn’t have obvious answers.

Consider:

Is that support or shortcut?

It depends.

If the learning goal is evaluating claims, maybe that’s useful support.

If the learning goal is generating an original argument, AI may have just done the hardest part.

That ambiguity is where the learning happens.

Students have to think about the purpose of the assignment, the skill being practiced, and their own level of understanding.

In other words, they have to think about their thinking.

Take an assignment you already use and create six to eight possible AI uses for it.

Mix together clearly helpful uses, clearly excessive uses, and a few debatable ones.

Then ask students to sort them and defend their decisions using this prompt:

You can finish by asking students to write one personal rule:

Those answers may be more useful than any AI policy poster we could hang on the wall.

Students are going to face hundreds of small decisions about AI use.

Most of those decisions will happen when no teacher is standing nearby.

So the long-term goal can’t simply be compliance with a list of approved and prohibited uses. Students need enough metacognitive awareness to recognize when a tool is helping them learn and when it is quietly replacing the struggle they need.

Spot the Shortcut gives them practice making that judgment while the stakes are low.

And that may be one of the most important AI skills we can teach: knowing when getting help is useful, and when the shortcut is taking you around the learning instead of toward it.

Want to see more? Check out the Learning Forge One-Pager for more ideas from the upcoming book.

After working with this year’s cohort of Kenan Fellows one last time this month, I’m gearing up to present the Learning Forge at the Association for Middle Level Education’s annual conference in Nashville in November. If you’ll be at AMLE26, I’d love to connect!

November’s newsletter will include a shout out to the folks I’m most thankful for this year.

Paul (and the Codium Educational Consulting team)

P.S.

Don’t forget to email me with examples of how you’ve used the tools and strategies that I’ve shared.