Teachomatic what to automate, and what not to

Differentiating for a Mixed Class

If you are going to use these tools for one thing, use them for this.

Differentiation is the task teachers most consistently intend to do and most consistently skip, for the simple reason that producing three versions of a text takes three times as long as producing one. That is mechanical labour, the thinking is already done, and it is where the measured time saving actually comes from.

For a workplace-oriented comparison outside education, the explanation offers another way to look at measurement, workload, behaviour, or accountability.

It is also the place where the output goes wrong in ways that are easy to miss, because a levelled text always looks fine.

What to ask for

Be specific. "Make this easier" produces something shorter and vaguer, which is not the same thing.

For an external perspective on lesson planning, teacher workload, and classroom technology, see Google for Education.

Reading level, named. Give a target — a year group, a reading age, a readability band your school uses. Then check it rather than trusting it, because the score measures sentence and word length rather than difficulty.

Sentence length and clause structure, separately from vocabulary. These are different levers and confusing them is the main cause of bad output. A text can have simple words and impossible sentences.

Vocabulary handling, specified. There are three options and you have to choose: replace the hard word, keep it and gloss it, or keep it and define it in place. For subject-specific terminology the answer is almost always keep it. A geography text that says "hot dry area" instead of "arid" has made the text easier and the learning worse, because the word is part of what is being taught.

Scaffolding, named. Sentence starters, a glossary, a worked example, questions broken into steps. Ask for the specific one you want.

The same content. Say this explicitly, because it is the thing that quietly slips.

The four failures to check for

Content shrinkage. The simplified version has lost an idea, not just simplified it. Compare the versions for what is in them before looking at how they read. This is the most common failure and the most damaging, because a student on the lower level ends up learning less rather than learning the same thing differently.

Vocabulary stripping. The subject terminology has been simplified away. Check specifically that the words you intend them to acquire survived.

False simplification. Shorter sentences, same conceptual load. Chopping a complex sentence into three does not make the idea easier; sometimes it makes it harder, because the connectives that showed the relationship are gone.

Register collapse. The lower-level version reads as though written for a younger child. A fourteen-year-old who needs a simpler text is not a nine-year-old, and they will notice immediately. This is the failure students actually complain about.

Levelling is not the only axis

Reading level dominates the conversation because it is the easiest thing to vary. Other axes matter more for some students and are just as cheap to produce:

Amount — fewer questions, same difficulty. Often the right answer for a student who is capable but slow.

Scaffolding, at the same reading level — same text, plus a structure to work through.

Entry point — the same concept approached through a different example, closer to what a particular student already has.

Output format — the same understanding demonstrated by a diagram, a table, or a spoken answer rather than a paragraph.

Only the first is about text difficulty at all. It is worth asking which one a given student actually needs before automatically producing an easier version — a decision that depends on knowing the particular child, which is the part that does not hand over.

The check that takes two minutes

Print the versions side by side and read the lower one first.

Does it contain every idea the top version contains? If not, fix it or accept deliberately that you are teaching less.

Does the subject vocabulary survive?

Would you be happy for this student to be seen holding this sheet? The dignity test, and the one that catches register collapse.

Could a student move up a level mid-lesson? If the versions diverge too much they cannot, and differentiation has become tracking.

That last question is the one worth taking seriously. Cheap levelled materials make it very easy to produce a permanent three-tier classroom without ever deciding to. The materials should be a support a student can step off, not a lane they are placed in.

The genuinely good news

This is the clearest case on the whole site of a task that machines automate and that teachers were skipping for lack of time — not because it was unimportant, but because the labour was disproportionate.

A class where every student has a text they can actually read is a better class. That was always true, and it used to cost an evening per lesson. It now costs about ten minutes plus a careful read, and the careful read is not optional.

Unlike most claims in this area, this one does not depend on contested evidence. You can see whether it worked by watching whether they read it.

The short version