AI in the Classroom

In this episode we explore a problem with AI tutors that may have very little to do with the AI itself.

Drawing on a recent conversation between Rick Hess and Michael Horn, we look at the “5% problem”—the gap between how widely an education technology is deployed and how many students actually engage with it meaningfully. 

We also explore why gamification may not solve the engagement problem, how motivation complicates research about AI and learning, and a practical way educators can test the assumptions behind new AI tools before rolling them out at scale.

Topics covered:

• Why student motivation may be the overlooked challenge facing AI tutors
• The “5% problem” in education technology
• The limits of points, badges, and gamification as substitutes for authentic engagement
• Justin Reich’s argument that recommendations for classroom AI should be treated as hypotheses
• What teachers and district leaders should ask before deciding an AI product “works”

AI in the Classroom Daily helps educators make sense of AI without the hype.

Hosted by Dan Cogan-Drew, co-founder and Chief Academic Officer at Newsela.

Source:

https://www.edweek.org/technology/opinion-the-big-flaw-with-ed-tech-tutoring-isnt-technological/2026/09

In this episode we explore a new randomized controlled study on what happens when people get help from AI, and then have to work without it.

In this episode we explore:

• What researchers found when they gave learners AI assistance and then took it away
• Why AI may change a student’s perception of how difficult a task should feel
• The connection between AI, cognitive offloading, and productive struggle
• Why the design of the study itself raises important questions about its conclusions
• How relevance and authentic assessment might change the way students respond to AI
• What this research suggests teachers should consider when designing assignments

AI in the Classroom Daily helps educators make sense of AI without the hype.

Hosted by Dan Cogan-Drew, co-founder and Chief Academic Officer at Newsela.

Source:

https://arxiv.org/pdf/2604.04721

In this episode we explore what happens when the texts students learn to read are increasingly written by AI.

What expertise does an educator need to judge whether the resulting text is actually good for students?

Topics covered:

• Why AI-generated reading passages are becoming attractive to educators
• The tension between practicing reading skills and building knowledge
• Whether “learning to read” and “reading to learn” can be separated
• What happens when AI rewrites text
• Why hitting a target Lexile isn’t necessarily good
• What can disappear when a text is simplified

Sources:

https://www.edweek.org/teaching-learning/more-kids-are-learning-to-read-on-ai-generated-text-what-are-the-pros-and-cons/2026/08

https://help.newsela.com/en/articles/13656260-newsela-s-approach-to-leveling-nonfiction-texts

In this episode we explore four stories shaping the conversation around AI in schools.

Florida has approved new AI requirements for schools that raise some surprisingly complicated questions about parental consent, student access, and even what counts as an “AI tool” in the first place.

We also revisit a provocative argument from MIT’s Justin Reich around expert AI guidance.

Then we look at the growing effort to independently evaluate AI products being used in schools.

Finally, we explore a new wrinkle in the battle over phones and devices in schools.

Topics covered:

• Florida’s new rules for AI use in schools
• Why parental consent could become a major AI policy issue
• The growing challenge of defining what actually counts as an “AI tool”
• Justin Reich’s warning about today’s AI best practices
• New efforts to independently evaluate AI learning tools
• What districts should look for when judging AI products
• AI smart glasses and the next challenge for school device policies

Sources:

https://www.cbsnews.com/miami/news/florida-board-education-ai-rules-k12-colleges-2027/

https://publicaisummit.org/people/justin-reich/

https://www.edweek.org/technology/new-project-identifies-strengths-and-weaknesses-of-a-collection-of-ai-learning-tools/2026/09

https://ailearningtour.instructionpartners.org/

https://www.edweek.org/technology/districts-expand-cellphone-bans-to-target-ai-smart-glasses-why-it-matters/2026/09

In this episode we explore the question: Does AI help or hurt learning?

Drawing on a keynote from Princeton computer science professor Arvind Narayanan at the Public AI Summit for Educators, we examine why that may actually be the wrong question, and why the conditions surrounding AI use may matter more than the technology itself.

Topics covered include:

• Why “Does AI help or hurt learning?” may be the wrong starting point
• The difference between essential skills and incidental skills
• When offloading work to AI might free students to focus on deeper thinking
• The calculator analogy
• The complicated relationship between AI, writing, and assessment
• Why schools may need a broader portfolio of ways to evaluate student understanding
• The growing disconnect between how students are expected to work in school and how AI is being used in the workplace

Source:

https://publicaisummit.org/keynote/does-ai-help-or-hurt-learning-wrong-question/

In this episode we explore a provocative idea from MIT professor Justin Reich: when it comes to AI in education, many of today’s “best practices” may not actually be best practices at all.

Drawing on Reich’s keynote at the Public AI Summit, we examine the research on AI and learning. We also explore why seemingly sensible AI guidance can age quickly, what Google and MIT’s own uncertainty about teaching people to work with AI can teach K–12 educators, and where to find evidence during this period of rapid change.

Topics covered:

• Why AI “best practices” in education may be premature
• What educators should make of gaps in current AI research
• What teachers should do with current AI practices
• Why state-level AI guidance deserves thoughtful scrutiny
• What Google and MIT’s uncertainty about AI can teach K–12 schools
• Lessons from the history of media literacy and lateral reading

Sources:

https://www.aiedu.org/educator-resources

https://datanutrition.org/

https://publicaisummit.org/keynote/the-homework-machine/

In this episode we explore what can get lost when asking a person turns into asking a machine, and why the ability to ask other people for help may be a skill schools need to deliberately protect and practice.

Drawing on recent work from Pamela Cantor and Julia Freeland Fisher, we examine why young people may avoid reaching out, and what happens when AI becomes an easier, frictionless alternative.

Topics covered:

• Why students may choose AI over asking another person for help
• The misconception that asking for help is a burden
• Julia Freeland Fisher’s concept of “pro-social AI”
• What human relationships provide that AI interactions cannot
• Why “weak ties” and social networks matter for students
• How AI could affect students’ ability to seek help and build relationships
• A classroom routine for practicing how to identify and approach human experts

Sources:

https://biologyofbecoming.substack.com/p/trauma-dumping-is-what-friendship

https://substack.com/inbox/post/214866089

https://www.the74million.org/article/survey-young-people-turn-to-ai-to-be-their-real-unfiltered-selves/

In this episode we explore what authentic student engagement can look like in the age of AI.

Stories of extraordinary project-based learning can be inspiring but can also feel out of reach for the average teacher trying to make tomorrow’s lesson more meaningful.

We look at a different model from Notre Dame’s God and the Good Life course and think about how we design learning experiences students actually need to inhabit for themselves.

Topics covered:

• What AI is exposing about some schoolwork
• The limits of using project-based learning as a model for everyday classrooms
• The difference between redesigning an assignment and redesigning an entire school
• How teachers can increase engagement within the constraints of their existing classrooms
• What authentic student work might look like in the age of AI

Sources:

https://www.the74million.org/article/schools-have-a-motivation-problem-ai-is-making-it-impossible-to-ignore/

https://philosophy.nd.edu/courses/1st-courses-in-philosophy/god-and-the-good-life-an-introduction-to-philosophy/

In this episode, we explore three stories that look encouraging on the surface, but get much more complicated when you look closely.

A school district says AI helped students become better critical thinkers.

SchoolAI is making new commitments around student privacy. 

And a new national framework could change how future teachers are prepared to use AI before they ever enter the classroom.

In this episode, we explore:

• The surprising results of a two-year classroom AI study
• Why “AI improved critical thinking” may not mean what it sounds like
• A new development in the fight over student data and AI
• The privacy problem that contracts alone may not be able to solve
• Why teacher preparation programs are suddenly rethinking AI
• The skills future teachers may need that weren’t part of teacher training just a few years ago

Sources:

https://www.eschoolnews.com/digital-learning/2026/09/09/aacte-releases-national-framework-on-ai-in-educator-preparation/

https://www.edweek.org/technology/microsoft-agrees-to-new-student-privacy-protections-for-ai-how-ironclad-are-they/2026/09

https://www.k12dive.com/news/utah-district-reports-critical-thinking-boost-amid-ai-deployment/829919/

In this episode we explore the question, what if AI were designed to strengthen human relationships rather than replace them?

We look at Prosocial Possibilities, an early-stage study from the Clayton Christensen Institute examining how AI and technology might help students identify, reach out to, and build stronger relationships with mentors, counselors, alumni, and other people in their networks.

Topics covered:

• The Clayton Christensen Institute’s Prosocial Possibilities research
• How AI affects students’ habits of asking other people for help
• The difference between AI that replaces relationships and AI that facilitates them
• What the student experience with Khanmigo can teach us about AI tutoring
• How Uprooted Academy used technology to help students activate their support networks
• How Overgrad explored reducing the anxiety and friction of professional networking
• Why some kinds of “friction” in learning are productive—and others may not be

Source:

https://www.christenseninstitute.org/publication/prosocial-possibilities/