IS CHATGPT THE WORLD’S MOST INFLUENTIAL TEACHER?

I recently read an article by Philippa Hardman who argues that while most people use ChatGPT for learning, much of that “learning” is shallow. It feels productive but often doesn’t stick. GenAI makes things look easy (e.g., clean summaries and polished drafts) but that can trick us into thinking we’ve learned when we’ve only skimmed the surface.

Why this happens

Hardman identifies three traps that create an “illusion of learning”:

• Fluency bias: Information looks familiar and smooth, but you don’t really remember it.
• Performance vs. learning: You get through practice quickly, but it doesn’t transfer to new contexts.
• Over-scaffolding: Too much step-by-step help means you don’t build independent problem-solving skills.

The fix

She proposes ways to push learning beyond easy answers and into durable knowledge:

• Recall without cues (blank page test).
• Explain concepts to someone else.
• Mix problem types so you have to adapt.
• Add constraints (timed tasks, rubrics).
• Space out review over time and check a week later.

My takeaway

This connects strongly to how we design instruction. Teachers can redesign assignments so they build real learning, not just polished outputs. Here are concrete moves that align with Hardman’s principles:

1. Add pre-task struggle: Have students attempt a problem or question before they’re allowed to ask AI for help.
2. Require retrieval: Instead of letting students copy summaries, ask them to recall key points from memory in writing or in discussion.
3. Use explain-to-others tasks: Assign peer teaching, where students must explain a concept to a classmate or in a short video.
4. Incorporate interleaving: Mix different problem types or topics in the same assignment so students have to select strategies, not just repeat a pattern.
5. Design authentic performance tasks: Replace short-answer questions with case studies, debates, or projects where students must apply knowledge in context.
6. Leverage timed or rubric-based work: Set time limits or require students to meet specific rubric criteria.
7. Ask for process evidence: Require students to show their draft steps, thought process, or even their AI interactions.
8. Schedule spaced reviews: Build in checkpoints where students revisit and use earlier content instead of moving on and never returning.

Assignments that reflect these practices ensure AI isn’t just a shortcut generator, but a tool to help build memory, problem-solving, and transfer skills in students.