AI in education: separating the hype from the evidence

In this episode, Svenia Busson sits down with Wess Trabelsi, a Tech Integration Specialist at Ulster BOCES in New York, where he supports eight rural school districts. Wess is neither a cheerleader nor a doomsayer when it comes to AI in education, he's something rarer: an evidence-driven practitioner who actually read the research.

Wess shares his deep dive into the science (or lack thereof) behind AI in K-12. After reviewing over 100 studies, he found that the vast majority are noise, glorified surveys, opinion pieces, and what he calls "dead horse studies" that prove the obvious. His findings closely mirror those of the Stanford AI Hub for Education's newly released 2026 Review, which started with 800 studies and kept only 20 with strong causal evidence, and found zero conducted in U.S. K-12 settings.

Together, Svenia and Wess unpack the two most significant studies to date: the Harvard RCT showing a custom AI tutor significantly helped motivated physics students, and the landmark Wharton/Turkey study showing that AI-assisted practice gains completely disappeared when AI was removed at test time. Neither provides a clear playbook for the average classroom.

But the conversation goes much deeper. Wess makes the case that AI didn't create a problem in educatio, it merely exposed one that was already there, putting "the final nail in the coffin of the traditional model." He advocates for process-based and project-based learning (citing the inspiring model of High Tech High in California), and for rethinking assessment entirely, away from written essays as proxies for thinking, toward conversation, video, and authentic real-world problem-solving through work-based learning.

If you've been overwhelmed by AI headlines in education and wished someone would just tell you what the evidence actually says, this episode is for you.

Resources mentioned in this episode:

• 📄 The Evidence Base on AI in K-12: A 2026 Review — Stanford AI Hub for Education (PDF)
• ✍️ The Good, the Bad, and the Ugly Science of AI in Education — Wess Trabelsi (Substack)
• 🎓 High Tech High — San Diego, California (school that does PBL)
• 📖 Superintelligence by Nick Bostrom · Works by Ray Kurzweil

Substack writers recommended by Wess:

• Mike Taubman — AI Waypoints
• Stephen Fitzpatrick — Teaching in the Age of AI
• Mike Kentz — How We Frame Machines