Above, PathoScan student creators Noah Choucair, Maanya Lalithraj, and Chloe Jreije present their prototype at the 2026 ISTELive conference in Orlando, Florida.
On the first day of a graduate class at the University of Pennsylvania, I read the most permissive AI policy I had ever seen. My professor, Jeffrey Babin, let us use any tool we wanted for any task we wanted. Two conditions: Disclose every use and submit your full prompt history with the work. He had drafted the policy itself with ChatGPT and said so in the syllabus, modeling the disclosure before the first class even met.
My initial thought was, Finally! Someone who gets it. For years I had been building the same kinds of requirements into my own financial literacy classrooms in New Jersey—making students present their work and defend it out loud—because that was the part no chatbot could do for them. It was great to see a professor doing it at the graduate level.
That class became the hardest I worked in my entire program. What I underestimated was everything my professor built around that permission. One day, he gave our groups a single hour to build a presentation about a company, and at the end of the hour, one of us from each group had to stand and deliver it while he and his teaching assistant graded us in real time. Before the final 20-page group paper was due, we had to present to the room the company we had chosen, our plan for organizing the paper, and how we would split the work. Classmates scored us on a Google form, and we received live feedback. On the paper itself, we had to declare who wrote each section and submit every AI prompt alongside the final output.
There was nowhere to hide. Every time I reached for AI, I had to account for the prompt and what I did with the answer. The tool sat right there, fully allowed, yet I had never been so aware of my own thinking.
That same semester, I took another class where no AI was allowed on anything we submitted, not even the discussion posts. I emailed the professor respectfully to make the case for AI as a collaborator, thought partner, and research assistant. She held to her policy.
In the class that banned AI, my energy went to that argument. In the class that welcomed AI, my energy went to the work. The rule decided where my mind went.
We spend a lot of energy counting the hours our kids are on screens. The count says very little on its own.
I think about that contrast every time I hear the phrase “digital wellness.” Most of our energy right now goes to subtraction. We ban the phones, cap the screen time, run student work through detectors. Some of that is necessary, and a phone buzzing through 3rd period is a real problem. But removing the distraction is only half the job. The other half is building work a student actually wants to think through.
I lead a program at the International Society for Transforming Education called GenerationAI, and this spring I coached the winners of our AI Innovator Challenge. One of those teams is comprised of three high schoolers from the United Arab Emirates, who built an app called PathoScan. They trained an AI model on roughly 7,000 real tissue images to recognize colorectal cancer, and they built the first working version in a week. When they tested it on new images it had never been trained on, it was 98.7 percent accurate, flagging all but two of the 223 samples that had cancer. PathoScan does not yet have FDA approval, and the students involved are teenagers, so the fair word is prototype. The truer word is astonishing.
We spend a lot of energy counting the hours our kids are on screens. The count says very little on its own. One hour disappears into doomscrolling. Another goes toward building something that might help a stranger live. The clock cannot tell you which.
The students we are anxious about today are the ones who will shape what comes next. What they learn to do with AI now is what all of us will live with later. So teach them to build with it. Measure a school by what its students make, and by whether they know when to lean on the machine and when to trust their own judgment. The fight over screen time will read like the first nervous question of a much larger conversation.
Most students will never build a cancer detector, and they do not need to. However, the same idea works in an ordinary classroom. A traditional history paper becomes a documentary the class will screen or a debate students have to argue in real time. A science fair project gets graded on the prompt thread as much as the poster. The tool stops being a shortcut the moment the work asks a human to show up.
The PathoScan developers spent their screen time on cancer detection. My students spent theirs building projects they had to defend in front of the room. I spent mine designing training to help teachers across the country use these tools. Doomscroll or cancer detector, it is all screen time. What matters is where it leads. Our job as educators for the next decade is to hand students work big enough to deserve their full attention.
