All issues
A ScholarStack Newsletter

The AI-Native Classroom

Applied AI in education — grounded in real classrooms, backed by emerging research.

Issue #1 · June 2026

Welcome to the first issue of The AI-Native Classroom — a monthly newsletter for the people building the future of teaching and learning with AI. Each issue: the three things you need to know, research worth your time, the best reads from across the field, and what's coming up. No hype, no fluff. Let's dig in.

📌Three things to know

The signal from the noise this month.

  1. 01

    The largest US public university system just set a systemwide AI policy. SUNY will require all 64 campuses to adopt AI guidelines by December 31, 2026 — covering data privacy, AI literacy in general education, and procurement safeguards. One of the first coordinated, system-level AI governance frameworks in US higher ed.

    Source: Inside Higher Ed · May 4, 2026

  2. 02

    The federal government is now funding AI in higher ed — to the tune of $169M. The US Department of Education awarded $169 million through the FIPSE program, with a major focus on responsible AI to improve teaching and student success, drawing a historic number of applications.

    Source: US Department of Education · Jan 5, 2026

  3. 03

    Adoption is near-universal — but readiness isn't. Student AI use has hit 95%, yet a new global survey found only one in three universities have a clear AI strategy and fewer than one in five have governance structures. The gap between using AI and using it well is the story of 2026.

    Source: IREX / Development Gateway · May 7, 2026

🔬Research worth your time

Two US classroom experiments that reframe the whole debate.

Scientific Reports (Nature) · June 2025

AI Tutoring Outperforms Active Learning in a Harvard RCT

Greg Kestin, Kelly Miller, and colleagues ran a randomized controlled trial with 194 undergraduates in Harvard's largest introductory physics course. Each week, students were randomly assigned to either in-class active learning or a custom AI tutor ("PS2 Pal") built with expert-authored scaffolds, step-by-step reasoning prompts, and guardrails against hallucination. Students using the AI tutor achieved roughly twice the learning gains of the active-learning group (effect size 0.73–1.3 SD), and reported higher engagement and motivation. The tutor revealed only one step at a time and prompted students to think before showing answers — mirroring the pedagogy of the in-person class.

Kestin et al. (2025), Scientific Reports 15, Article 17458. doi.org/10.1038/s41598-025-97652-6

◆ Why it matters

This is the first rigorous RCT to show a custom AI tutor beating best-practice active learning — not a bad lecture, a good class. The key variable was pedagogical design: the AI scaffolded thinking rather than just answering.

Proceedings of the National Academy of Sciences · June 2025

Generative AI Without Guardrails Can Harm Learning

A University of Pennsylvania (Wharton) team led by Hamsa Bastani ran a field experiment with nearly 1,000 high school math students across three groups: "GPT Base" (standard ChatGPT-4), "GPT Tutor" (safeguarded to give hints, not answers), and a no-AI control. Both AI groups improved during practice (48% and 127%). But on a follow-up exam without AI, GPT Base users scored 17% worse than the control — the unguided tool had become a crutch. GPT Tutor users kept their gains. The guardrails preserved actual learning.

Bastani et al. (2025), PNAS 122(26), e2422633122. doi.org/10.1073/pnas.2422633122

◆ Why it matters

Read together, these studies settle the framing: the question isn't "AI vs. no AI" — it's "designed AI vs. unguided AI." Same technology, opposite outcomes, depending entirely on pedagogical design.

📚AI in education reads

Three frameworks you can put to work this month.

Times Higher Education
AI and Assessment Redesign: A Four-Step Process

If AI means institutions can no longer assure the integrity of individual assessments, the focus must shift to assuring the integrity of overall awards. A practical framework for faculty rethinking how they evaluate learning.

Read more ›
University of Iowa · Center for Teaching
The AI Assessment Scale: A Framework for Designing AI-Aware Assignments

A five-level scale — from "No AI" to "AI Exploration" — that gives instructors a shared vocabulary for specifying exactly how much AI use each assignment allows. A practical tool for redesigning tasks and communicating expectations to students, rather than policing them.

Read more ›
Every Learner Everywhere
AI Literacy Frameworks for Higher Education: A Faculty Guide

A curated map of the most useful AI-literacy frameworks (from EDUCAUSE, WCET, UNESCO, and others), organized by what each one helps students actually do. A practical starting point for choosing the framework that fits your course and discipline.

Read more ›
🎙️Worth your attention

From the ScholarStack team.

Learning with AI · Podcast

What does higher education need from AI that "off the shelf" genAI tools fail to provide?

In the first episode of Learning with AI, David Miller, Founder of ScholarStack and Professor at Carnegie Mellon University, shares the story behind the creation of ScholarStack and explains why he believes AI in higher education requires a fundamentally different approach.

The conversation explores student engagement, learning science, personalized learning, educational technology, and the opportunities AI creates for both educators and institutions.

Topics discussed

  • AI in Higher Education
  • Personalized Learning
  • Student Engagement
  • Educational Technology
  • Learning Science
  • The Future of Learning

Learning with AI is the ScholarStack podcast exploring how artificial intelligence is transforming teaching, learning, and higher education.

Learn more
Summer Course · Now Enrolling

Teaching in the age of AI

Understand how modern AI works and what it means for teaching and learning.

This free summer course is designed for educators who want to move beyond the AI hype and build practical understanding of today's rapidly evolving technology and its impact on higher education.

Over four weeks, participants will explore:

  • The foundations of modern AI systems and large language models (LLMs)
  • Prompting, retrieval (RAG), agents, evaluation, and AI workflows
  • The opportunities and limitations of AI in higher education
  • How AI can support teaching, assessment, curriculum design, and student success
  • What it feels like to learn inside an AI-native educational environment

By the end of the course, participants will leave with:

  • Firsthand experience using AI as a learner
  • New ideas for integrating AI into teaching and student support
  • Greater confidence discussing AI within their institution
  • A framework for thinking critically about the future of education in an AI-native world

No technical background required.