AI in the Classroom

What does the latest research actually tell us about how generative AI affects student learning?

In this episode we explore four different patterns of AI use in K–12 classrooms, and why the way AI is designed and implemented may matter more than the tool itself.

Topics covered:

• The difference between AI-assisted performance and durable learning
• What can happen when students have unrestricted access to tools like ChatGPT
• Why guardrails and guiding questions may help preserve student thinking
• How purpose-built AI can encourage reflection and metacognition
• How teachers can test whether learning transfers once AI is removed
• Why schools should focus on how AI is configured, not simply which AI tool they adopt

Source:

https://barker.institute/documents/188/Young_People_learning_and_generative_AI.pdf

In this episode we explore why asking “Does AI tutoring work?” may be the wrong question.

A new Stanford University research brief argues that AI tutoring isn’t a single model. Instead, it exists along a spectrum—from fully automated AI self-study to traditional human tutoring—with very different levels of research support behind each approach.

We break down six models of AI tutoring, examine what the evidence currently tells us about each, and consider what districts should look for before investing in AI-powered tutoring. 

Topics covered:

• The six different models of AI tutoring
• Why “AI tutoring” is too broad a category for evaluating effectiveness
• Which tutoring approaches have the strongest research support
• Why fully automated AI tutoring still has significant evidence gaps
• The role human tutors can play alongside AI
• Three questions district leaders should ask before adopting an AI tutoring model
• Why implementation may matter as much as the technology itself

Sources:

https://scale.stanford.edu/sites/default/files/AI%20Tutoring%20is%20Not%20a%20Monolith%20What%20We%20Actually%20Know.pdf

In this episode we look at the growing arms race around AI watermarking and why new methods for identifying AI-generated text may not provide the reliable detection solution schools are hoping for. We also unpack new survey data from Common Sense Media showing how high school students are actually using AI.

Topics covered:

• The emerging arms race between AI watermarks and watermark-removal tools
• Why watermarking may not solve schools’ AI detection problem
• What new survey data reveals about high school students’ AI use
• The gap between student AI adoption and teacher guidance
• Students’ concerns about AI, critical thinking, and losing opportunities to learn
• New U.S. Department of Education guidance on education technology
• The shift from measuring EdTech adoption to measuring learning outcomes
• Why districts still face major variation in AI policy, privacy, accessibility, and data protection

Sources:

https://www.edsurge.com/news/department-of-education-issues-long-awaited-edtech-guidance-for-states-and-districts

https://www.commonsensemedia.org/research/teens-in-the-ai-era-schoolwork-and-skills-that-matter

https://watermark.sabrina.dev/

In this episode we explore why AI policy alone may not be enough for schools navigating a technology that is changing faster than the traditional policy cycle.

We reflect on a conversation with a veteran high school English teacher who feels a growing gap between district AI guidance and the realities teachers are encountering in their classrooms. 

Topics covered:

• The growing gap between district AI policy and classroom reality
• Why static policies may struggle to keep pace with rapidly changing AI
• Teacher concerns about trust, student AI use, and academic work
• The isolation educators can feel when there’s no venue to discuss AI openly
• What schools can learn from bringing AI skeptics and advocates into the same conversation
• The case for standing AI councils that include teachers, students, administrators, families, and board members

Sources:

https://www.brookings.edu/articles/enhancing-student-agency-through-school-ai-councils/

https://www.edutopia.org/article/talking-technology-school-staff

https://theimportantwork.substack.com/p/what-does-it-mean-if-your-ai-position

In this episode we explore new research on AI-supported tutoring that suggests that giving students access to an AI tutor, even during structured school time, doesn’t necessarily mean they’ll use it as intended. 

Topics covered:

• What a two-year study of Khan Academy’s Khanmigo found about student achievement and engagement
• Why structured practice may matter more than conversational AI tutoring
• What another study tells us about pairing human support with AI learning platforms
• Three qualities that distinguish effective human tutors
• The “reach-out problem”
• Why student motivation, confidence, and relationships matter for tutoring
• How AI tutors may need to evolve if they’re going to approach the impact of high-dosage human tutoring
  

Sources:

https://edworkingpapers.com/sites/default/files/ai26-1451.pdf

https://www.linkedin.com/events/7487956708225433600/

https://www.nber.org/papers/w35620#fromrss

In this episode we explore how AI can help school leaders reclaim time, reduce administrative burden, and focus more of their attention on the work that matters most: supporting teachers, students, and learning.

For this Teacher Tuesday conversation, we sit down with Carl Johnson, Associate Director at the New England Association of Schools and Colleges (NEASC) and a former school principal. 

We explore when should students use AI, and what happens when educators don’t have enough expertise to evaluate an AI-generated answer.

Topics covered:

• How AI can help school leaders reduce administrative workload
• Why educators may already have many of the skills needed to work effectively with AI
• Practical AI workflows for meetings, brainstorming, classroom observations, and follow-up
• Why human judgment and “critical distance” still matter when working with AI
• The importance of transparency when school leaders use AI
• What schools can learn from the adoption of smartphones and social media

In this episode we explore how competitive academic environments may be shaping the way students use AI, and why some students may feel they have little choice but to use it.

Can the way we grade, rank, and define success unintentionally turn AI from a useful option into a perceived necessity?

Topics covered:

• How social comparison can influence students’ AI use
• Why competitive students may experience AI-specific FOMO
• The difference between strategic AI use and anxiety-driven reliance
• How grading curves, class rank, and college admissions pressure can affect AI adoption
• Why students may use AI even when they have negative feelings about it
• How teachers can make expectations around AI use more explicit
• The role of AI-free assignments, transparency, citation, and reflection

Source:

https://arxiv.org/abs/2606.03560

In this episode we explore three AI-in-education stories that point to a growing challenge for schools: districts are investing in AI faster than researchers, policymakers, and educators can determine what actually works.

Topics covered:

• Why districts still lack clear guidance for evaluating AI tools
• What current research suggests about AI, cognitive offloading, and student learning
• What San Antonio ISD’s mixed results tell us about pilots, causation, and implementation
• Why positive student outcomes alone may not be enough to prove an AI product works
• Anthropic’s introduction of machine-readable watermarks for Claude-generated text
• How transparency about AI involvement could change assessment and academic integrity

Sources:

https://stateline.org/2026/08/12/schools-spend-billions-on-ai-but-struggle-to-figure-out-whats-worth-it/

https://ies.ed.gov/learn/blog/ai-k-12-education-good-bad-and-guardrails-consider

https://www.yahoo.com/news/us/articles/saisd-put-ai-middle-schools-100000048.html

In this episode we explore whether school districts should be more willing to say no to new education technology, especially generative AI.

We are joined by Claude for a conversation about the research paper Refusing Educational Technology: Artificial Intelligence, Inequity, and the Problem of Critical Optimism. The paper challenges a familiar stance in education: acknowledging the risks of new technology while assuming that, with the right guardrails, schools should ultimately find a way to adopt it.

Topics covered:

• Why “critical optimism” may still tilt schools toward technology adoption
• When saying no to an AI or EdTech product can be a legitimate district decision
• The difference between technology that supports teacher agency and technology that asks educators to step aside
• How districts can define success, failure, and acceptable outcomes before launching a pilot
• Why vendor usage data shouldn’t become the default measure of whether an implementation is working
• How districts can prevent optional tools from slowly becoming prescribed implementation models
  

Source:

https://journals.sagepub.com/doi/abs/10.3102/0013189X261466234

In this episode, we explore how one school district is responding to AI not as a passing trend, but as a rapidly evolving reality for students, teachers, and school leaders.

We are joined by Alicia Kersten, Principal of Somerville High School, and Jason Behrens, Innovation Project Specialist for Somerville Public Schools, for a conversation about how Somerville is approaching AI from both the classroom and district perspectives.

Topics covered include:

• Why Somerville believes schools can’t simply avoid AI
• How student AI use is changing teaching, writing, and assessment
• Protecting critical thinking while allowing students to use new tools
• Preparing students to use AI in college, careers, and technical fields
• How teachers can use AI without adding another overwhelming initiative
• Bringing students, teachers, administrators, and families into the conversation
• The practical tension between innovation, budgets, priorities, and implementation

Source:

https://www.aasa.org/resources/resource/students-first-act