One reading, four proficiency levels, same discussion
The situation
You have a section where reading levels span several years. Assign the real text and half the room cannot engage; assign the simple one and the rest are bored and, reasonably, insulted. The goal is one shared discussion where everyone has actually read something they could handle.
Steps
-
Pick a text worth the trouble
Choose something where the ideas are worth discussing at every level. A text that is only interesting in its original phrasing will not survive this process — pick for conceptual substance, not prose style.
What you only learn by doing it: If the text's value is its language, do not level it. Teach it as-is with support. Leveling is for content, not for craft.
-
Generate the variants
Paste the text or a URL and generate levels. Diffit also produces vocabulary lists and comprehension questions alongside each variant.
What you only learn by doing it: The generated comprehension questions are mostly shallow recall. Keep the vocabulary support, rewrite the questions.
-
Audit what the simplification changed
Read each variant against the original with one question: did any claim get stronger? Restore hedges by hand. This is the step people skip and it is the one that protects the content.
What you only learn by doing it: Do this with a highlighter on paper the first few times. You will start to see the pattern in how these tools flatten uncertainty, and after that you can spot it quickly.
-
Give students a way into the original
Gemini Notebook (formerly NotebookLM)
Load the original text and let students query it in plain language while seeing the source sentence each answer came from. Students working at a lower level still meet the real text — they just get a bridge to it.
What you only learn by doing it: This matters more than the leveled variants. The variants let everyone participate; this step keeps the original in the room.
-
Run one discussion, not four
Everyone read the same content, so everyone joins the same conversation. Do not group students by variant — that recreates the tracking the whole exercise was meant to avoid.
What you only learn by doing it: Do not tell students which variant they received. It changes how they participate, and they will compare notes anyway.
Where this breaks down
The serious risk is not that simplified text is too easy. It is that simplification strips hedges and qualifications, turning 'the evidence suggests a modest association' into 'this causes that.' Students at the lowest level then learn a confidently wrong version of the content and carry it into the discussion. Read every generated variant against the original for claims that got stronger as the language got simpler.
Diffit targets K-12 grade bands, so output can read as childish to adults. Rewrite the register even when the level is right — adult learners notice immediately, and being talked down to does more damage than a hard text.