Stop Using AI to Level Down Students

13

Remember “Stay gold, Ponyboy, stay gold”? It hits you right in the chest. It matters.

Now try this: “Keep being good and special, Ponyboy.”

Flat. Boring. Wrong. Most English teachers would call the second version an atrocity. It strips the soul out of S.E. Hinton’s masterpiece. Yet this is exactly what happens when educators use AI to level down text for struggling readers. Instead of scaffolding students up to complex literature, we dumb it down. We sacrifice depth for convenience. And students lose the very thing that makes reading worth it.

The Problem With Leveled Reading History

The idea that kids should only read at their “instructional level” has been around since the colonial era. Timothy Shanahan breaks this down well in his work Leveled Reading: Leveled Lives. He argues, much like Natalie Wexler did in The Knowledge Gap, that this practice limits cognitive growth.

The logic used to be simple. If a seventh grader reads like a third grader, give them third-grade books. Programs like Fountas & Pinnell or Accelerated Reader cemented this in schools. It spread from elementary interventions to general high school classrooms. The assumption was safety. Keep students in their zone of comfort.

Decades of research, particularly regarding the Science of Reading, prove this wrong. Comfort zones don’t build brains. Complex texts do. But only when paired with skilled teaching. Students need chunking. Vocabulary support. Discussion. Background knowledge building. They need a teacher guiding them through the density, not a robot summarizing it away.

Why Teachers Turn to AI Tools Like Diffit

Two years ago, my district in New Jersey went all-in on a new curriculum. Complex, grade-level texts. The pushback was immediate.

“It’s too hard,” teachers said. “Our dyslexic and ADHD students won’t make it through Shakespeare or All Quiet on the Western Front.”

I heard this from eighth-grade teachers terrified that their general education students couldn’t handle Remarque’s anti-war classic. That was the moment Diffit entered the chat.

Diffit markets itself as an AI tool for creating “just-right” instructional materials. Basically, it levels text. A respected Learning Disabilities Teaching Consultant (LDTC in my district suggested it. She told me to run chapters of the novel through the tool to generate third-grade-level summaries for special education students.

I hesitated. I knew the research. But she was a veteran. She knew disabilities better than I did. So I trusted her. I used the tool. I gave a dyslexic student with ADHD a shortcut.

It was a mistake.

What Happens When You Remove the Struggle

The results were lackluster. The student waited for answers. She showed zero interest. She didn’t remember content between sessions. Her special education teacher saw no improvement in class participation.

Why? Because learning requires effortful thinking. When you remove the cognitive load, you remove the learning.

I switched tactics. I went back to the real text. We mapped characters. We discussed themes. We explored vocabulary. We used spaced retrieval practice to lock information into long-term memory.

The change was instant.

Our discussions on Paul Bäumer’s trauma were profound. This student, who many had written off, began joining peer discussions in the general classroom. Three years later, those same skeptical teachers see All Quiet on the Western Front as a student favorite. Why? Because middle schoolers crave maturity. They want real topics. They don’t need watered-down summaries. They need access.

How to Use AI for Scaffolding Instead of Dumbing Down

The fault isn’t the AI tool. It’s how we use it. AI is powerful. It can be Mephistopheles to our Faustus: a helpful servant or a dangerous master. We must avoid the latter.

Don’t let AI replace student thinking. Let it scaffold it. Here is how to use tools like Diffit responsibly:

  • Identify friction points: Have AI flag syntactically complex or dense passages. Use those for targeted vocabulary and syntax instruction.
  • Chunk, don’t change: Use AI to break text into manageable phrases. Use slashes / to mark breath points. This builds fluency without altering meaning.
  • Select fluency material: Ask AI to identify passages suitable for repeated reading practice.
  • Model think-alouds: Struggle to verbalize your thought process? Have AI generate a script for modeling how an expert reader approaches the text.
  • Create word sets: Generate lists for deep morphological study. Analyze connotations. Study roots.

“Student success with challenging text… is part of the larger conversation about the quality tech tools we use.”

Equity means access. It doesn’t mean lowering the bar. It means lifting the student. Give every kid the chance to read the canon. Don’t give some the full experience and others a robotic summary.

AI supports creation. It supports access. It should never replace the work.

попередня статтяGulf of California Orcas Smashing Sunfish: New Insights on Orca Behavior
наступна статтяDid the 4-Strand Burau Representation Stay Faithful? The Final Proof