What a New Study Tells Us About AI and Learning

What Happens to Students When the Tool Is Gone

In this episode we explore what happens when students learn with AI or other cognitive supports, and then suddenly have to work without them. We look at a recent learning science study highlighted by Carl Hendrick in The Learning Dispatch, where students who had been using a simple planning tool saw their performance drop when that tool was removed.  Topics covered: * Why introducing an AI tool also means planning for its removal * The difference between scaffolds, accommodations, and persistent AI tools * What learning science suggests about tool dependence * Why “taking the tool away” can reveal whether students have actually built skill * Why teachers need control over when AI feedback is on, limited, or turned off * Why AI implementation should be judged not only by access and efficiency, but by whether students can still think and perform when the support is gone Sources: https://carlhendrick.substack.com/p/the-monthly-dispatch-whats-new-in-a9a https://link.springer.com/article/10.1186/s41235-026-00722-0

What a New Study Tells Us About AI and Learning

In this episode we explore a simple question for classrooms using AI: Should students think first and use ChatGPT later? We connect research to a seventh-grade poetry classroom where students drafted on paper before using AI-supported tools. Topics covered: * Why “better work with AI” is not the same as learning * The difference between performance and transferable skill * What happens when students go straight to ChatGPT for answers * Why “think first, AI later” may be a stronger classroom model * How teachers can structure AI use without banning it * Why assignment sequence matters as much as assignment policy * The risks of AI-assisted brainstorming at the very start of student work Sources: https://carlhendrick.substack.com/p/the-monthly-dispatch-whats-new-in-a9a  https://link.springer.com/article/10.1007/s10648-026-10118-7

The Friction That Teaches

In this episode we explore how AI tools can either preserve or undermine one of the most important parts of learning: productive struggle. We look at Student Achievement Partners’ recent piece on AI in literacy and math classrooms, with a focus on the caution that students may lose essential learning opportunities when AI does the thinking too early. Topics covered: * What Student Achievement Partners warns about AI in literacy and math classrooms * Why productive struggle matters for student learning * Manu Kapur’s research on productive failure * The difference between cognitive surrender and skipped thinking * How AI can either extend or eliminate learning effort * What teachers should look for when AI enters an assignment * How instructional coaches can observe AI use in classrooms * What district leaders should ask before approving AI tools Sources: https://learnwithsap.org/cross-content/ai-is-so-hot-right-now-opportunities-and-cautions-in-literacy-and-math-classrooms/ https://arch.kuleuven.be/studeren/tall/artikels/productive-failure-kapur.pdf/@@download/file/Productive%20Failure%20Kapur.pdf https://www.tandfonline.com/doi/abs/10.1080/07370000802212669 https://www.tandfonline.com/doi/full/10.1080/00461520.2016.1155457 https://bellwether.org/publications/productive-struggle/?activeTab=1

Is This the Better Way to Prove Student Authorship?

In this episode we explore a practical classroom response to one of the hardest problems AI has created for teachers: how to know whether a student actually understands the work they submitted. Rather than relying on unreliable AI detectors, we look at an emerging approach from educators like AJ Giuliani: using a student’s submitted paper to generate a short comprehension quiz that helps reveal whether the student truly knows the material. Topics covered: * Why AI detectors are often unreliable for judging student authorship * How quiz-based authentication can help teachers assess whether students know the work they submitted * AJ Giuliani’s classroom approach to AI-era authorship questions * Nick Potkolitsky’s distinction between “ownership” and “formation” * Why retention, transfer, and agency matter in student learning * What “generate before you delegate” means for student writing * How oral explanation, reflection, and defense of ideas can strengthen assessment * Why classroom bots may be useful for probing student understanding, not just catching misuse * Practical implications for teachers, instructional coaches, and district leaders Sources: https://ajjuliani.beehiiv.com/p/the-easiest-way-to-stop-ai-plagiarizing https://nickpotkalitsky.substack.com/p/is-ownership-the-best-metaphor-for

When the University Goes All In

In this episode we explore California State University’s large-scale AI licensing deal with OpenAI, and what it reveals about the tension between institutional policy, classroom practice, and pedagogical purpose. CSU set out to become the nation’s first “AI-powered university system,” offering ChatGPT enterprise access across its campuses. But a year later, many licenses went unused, faculty pushed back, and the deeper question became clear: providing access to AI is not the same as knowing what AI is for in teaching and learning. Topics covered: * CSU’s $17 million OpenAI agreement and what happened after rollout * Why enterprise AI access can be an administrative decision, but classroom AI use is a pedagogical one * Faculty resistance, unused licenses, and the limits of top-down implementation * The difference between AI as institutional branding and AI as a teaching tool * What K–12 leaders can learn from CSU’s rollout * How districts should think about AI licenses, adoption, teacher agency, and purpose * The risk of confusing workforce preparation with meaningful learning Source: https://www.nytimes.com/2026/06/01/magazine/ai-university-college-california.html