Sparked by an the work of Jonathan Haidt (author of “The Anxious Generation” among others), and fueled by the OECD’s finding that “Students who use computers very frequently at school do a lot worse in most learning outcomes”, a growing movement is pushing for less screen time in schools and even the complete abolishment of technology for educational purposes. Many experts (like the Candice Ogders Lab and Joan Ganz Cooney Center), however, note that it’s not the amount of time spent with technology, but rather the quality of the learning experiences that really makes a difference. Stay tuned for more about this in coming weeks as the debate continues.

Following closely on the heels of Google’s release of Nano Banana 2, OpenAI has now launched ChatGPT Images 2.0. Both bring image generation to a new level, minimizing the gibberish text that has been common in infographics and educational images until now. They unlock the ability for teachers to create amazing graphics to illustrate educational concepts, using simple prompting strategies. Check out next month’s AI-Supported Lesson for more information.

In a move that some are seeing as evidence that AI is forcing us to return to paper-and-pencil assessments, Princeton University is ending it’s 133-year-old tradition of relying on students to report their peers for violations of the honor code. The new code follows moves by Stanford in 2023 and many other universities in 2025, and will require human proctors for all in-person exams. While the change required a vote by the faculty, reports suggest that it was initiated by demand from the student body to rein in widespread cheating.

The story of AI hallucination is an old one, but even when these tools don’t confidently invent facts we know the value of keeping a human in the loop. Students need practice to learn the habit of always reviewing AI output–even small changes and suggestions–before submission. I do this in my own classroom by creating intentionally flawed materials and getting students to identify (and often verify) the errors.

No matter what subject or grade level you teach, there are ways that you can create output that has errors for students to find. Consider:

Generate these types of content using your favorite chatbot and a standard prompt, but be sure to ask the AI to make a specific number of errors and to list those errors at the end. Here are some other tips:

You can ask for factual errors, weak reasoning, bad evidence, biased wording, or correlation/causation confusion.

What’s great about this strategy is that is hits lots of instructional goals simultaneously. Students get to practice critical thinking while also learning that AI tools make mistakes. They become engaged in the struggle to catch the machine’s mistake while also developing a deeper understanding of the content. Best of all, it works just as well with students who are too young to be using AI tools directly but whom need to build the maturity and skepticism to be ready when they are.

The school year is winding down and I am going to be working with groups of educators who get the opportunity to learn before they start their summer break. I’ll be presenting at Plano ISD’s Inspire Summer Conference later this month in Texas. Then, I’ll be spending a week in June working with the new Kenan Fellows in the mountains of North Carolina.

In a year of big conferences, I’ll be giving a keynote presentation at the North Carolina Career Technical Education Conference, and presenting at the PACKed Beginning Teacher Conference, and Association for Middle Level Education Conference. Hope to see some of you at these events!

In April, I’ll share more of the Learning Forge here, giving you a sneak peak inside my latest book.

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.