
S2E365: Is AI Good for Classrooms? 6 Learning Science Principles It Can Help Deliver (A Former Teacher Explains)
Notas do episódio
Is AI going to help students learn better, or is it the worst thing we could put in a classroom? It’s one of the biggest debates in education right now, and Mackenzie makes the case that decades of learning science research point to practices a single teacher in front of a classroom simply can’t deliver at scale.
To dig into it, Mackenzie sits down with Sam DePalo, a former traditional teacher and Alpha guide with over 10 years of classroom experience. A Fulbright scholar with degrees in linguistics, literature, and mathematics, plus a postgraduate focus on neurodiversity and executive function, Sam taught high school math through Teach For America in a Title I public school (with a class of 47 students), then worked in private schools with much smaller class sizes. She saw the same limits in both.
In this conversation, Sam explains six learning science principles and why she believes an AI tutor (not a chatbot) can help deliver them, while freeing teachers to focus on motivating and connecting with kids.
In this episode, we discuss:
- Why the Traditional Model Falls Short: How the industrial-era classroom was built to educate the masses, and why Sam saw the same limits in a 47-student Title I class and in small private-school classrooms.
- One-to-One Instruction: Why individualized learning is the most powerful principle, and how an AI tutor can assess what a student knows and meet them at the right pace and level.
- Mastery-Based Learning: Why moving on before a student has mastered a concept leaves gaps, what a 90% mastery bar looks like, and how grade inflation and the bell curve blur what an “A” really means.
- Spaced Repetition and Retrieval Practice: How constantly revisiting past material, and being quizzed on it, beats cramming, and why a teacher with 25+ students can’t realistically build this in for every kid.
- The Zone of Proximal Development: How kids shut down when work is too hard or too easy, and why productive struggle (not gamified, easy learning) is the goal.
- Cognitive Load Theory: Sam’s “working memory whiteboard” explanation of why shaky math facts get in the way of learning algebra.
- The Teacher’s Role: What changes when AI takes on grading, data analysis, and differentiation, so adults can focus on motivation and connection.
Episode Timestamps:
00:00 - The great debate: Is AI good for classrooms or the worst thing for them?
01:37 - Sam’s background: from Fulbright scholar to Teach For America
04:16 - Why learning science doesn’t support the one-teacher model, even in small private schools
07:54 - Learning Science Principle 1: One-to-one instruction
09:08 - How an AI tutor (not a chatbot) makes individualized learning possible
12:36 - Learning Science Principle 2: Mastery-based learning
13:30 - Why Alpha sets the mastery bar at 90%
16:31 - Grade inflation, the bell curve, and the paper that got a 95% without being read
19:06 - Learning Science Principle 3: Spaced repetition and distributed practice
22:49 - Why cramming fails: the accounting class story
24:18 - Learning Science Principle 4: Retrieval practice
28:03 - Standardized assessments as feedback, not high-stakes judgment
30:00 - Hole-filling in action: how a student went from a 720 to a 790 on SAT math
32:12 - Learning Science Principle 5: The zone of proximal development
36:02 - Productive struggle vs. gamified learning
37:21 - Learning Science Principle 6: Cognitive load theory
41:19 - Why a motivated student matters most
43:13 - What AI frees teachers to do
Watch the Future of Education on YouTube: www.youtube.com/@future_of_education
Read MacKenzie’s Substack: www.futureofeducation.substack.com
Learn more about our schools: www.2hourlearning.com/schools
Learn more about Alpha School: www.alpha.school