Harvard Thinking Preserving learning in the age of AI shortcuts
137 snips
Feb 18, 2026 Ying Xu, a Harvard education researcher focused on AI in classrooms; Tina Grotzer, a cognitive scientist who studies metacognition; and Michael Brenner, a Google research scientist who studies large language models. They debate how to keep learning essential while using AI. They explore redesigning tasks, developmental differences, metacognition, tutoring’s social role, and guardrails to protect cognitive growth.
AI Snips
Chapters
Transcript
Episode notes
Learning Is Facts Plus The Ability To Learn
- Learning is twofold: the facts students learn and their ability to learn in the future.
- Critical thinking and metacognition are foundational capacities that schools must cultivate early.
Make Students Solve What AI Can’t
- Do raise the bar for student work when AI can solve standard problems.
- Ask students to invent unsolvable-by-AI problems and verify solutions to push deeper learning.
Homework Solved By Gemini Forced A Redesign
- Michael put his last homework into Gemini and it solved everything, so he changed his class.
- The class produced 600 problems chatbots couldn't solve and published a paper with student authors.