Before Comprehension: When a Child Cannot Enter the Text

When we talk about reading difficulties, we often ask a familiar question:

Does the child understand the passage?

We give them a text, ask them to read it, and then check their comprehension through questions.

But while working with children who have relatively weak Chinese foundations, I began to notice something more fundamental.

For some children, the problem starts before comprehension.

They are not yet struggling with the deeper meaning of the text.

They are struggling to enter the text at all.

A child may know the characters, but still not be able to read

During one learning session, I watched a Primary 5 child read a Chinese passage.

She could recognise some individual Chinese characters. She could also follow along when the text was read aloud.

But when she encountered a phrase such as:

“离放学还有两小时”

she did not naturally see the words inside the sentence.

She tried to select the part she did not understand, but instead of identifying a meaningful word or phrase, she separated characters in ways that did not correspond to the actual word boundaries.

That small interaction revealed something important.

The difficulty was not simply:

“She doesn't know the meaning of this word.”

There was an earlier problem:

She could not reliably see where the words were.

And if a child cannot identify the words in a sentence, asking whether she understands the sentence may already be asking a question too far downstream.

Reading can break long before the comprehension question

We often imagine reading as if the child moves directly from characters to understanding:

Characters → Words → Sentences → Meaning → Text

But each step depends on the one before it.

A child may recognise many characters and still struggle to combine them into words.

They may understand individual words but lose the sentence.

They may understand one sentence but fail to hold enough meaning in memory to connect it with the next.

By the time they reach the end of a paragraph, very little of the text may remain available to build a coherent understanding.

From the outside, this can look like a “comprehension problem.”

But comprehension may only be where the difficulty finally becomes visible.

The actual breakdown may have happened much earlier.

This led me to a different question:

Before asking whether a child understands a text, can they actually get into it?

“Entering the text” became a product problem

This observation began to change how we designed SmartRead.

Our earlier reading experience assumed too much.

We presented the passage, divided it into manageable sections, and provided tools such as AI explanations, vocabulary support and reading assistance.

These features were useful.

But for a weaker reader, we were still essentially saying:

Here is the text. Start reading. Ask for help when you need it.

That works only when the child already knows how to begin.

Some children do not.

So we started experimenting with a different approach.

Instead of immediately placing the child in front of the full passage, SmartRead first helps them establish enough context to enter it.

A short Story Trailer introduces the situation without revealing the whole story.

An AI Teacher Guide helps the child orient themselves to what is happening.

Then the child begins reading with some idea of the people, situation and question that matter.

The purpose is not to replace the text with pictures or explanations.

It is to make the text enterable.

The goal is always to return to the text

This distinction matters.

It would be easy to respond to a struggling reader by adding more and more support:

But support can become another way of avoiding reading.

SmartRead is not trying to make the Chinese passage unnecessary.

The opposite is true.

Every scaffold should eventually help the child return to the text with greater independence.

That means the learning path should not end with:

“The AI explained the passage to me.”

It should move toward:

“I can now read this passage myself.”

This is why we are increasingly thinking about SmartRead not simply as a collection of reading features, but as a sequence of support that can gradually step back.

This changes what AI should do

It also changes the role of AI.

If the goal were simply to answer questions about a passage, a general-purpose language model could already do much of the job.

A child could paste a sentence into an AI chatbot and ask:

“What does this mean?”

But education requires a different question:

What help does this learner need at this moment — and how much help should we give before stepping back?

Sometimes the right intervention may be an explanation.

Sometimes it may be a picture.

Sometimes it may be a simpler question.

Sometimes the system should not explain anything at all.

The challenge is not to maximise AI assistance.

It is to provide the minimum useful support that allows the learner to continue.

We are still learning

SmartRead is being built through repeated observation of children actually using it.

That process has repeatedly challenged our assumptions.

A feature that seems obvious when designing at a desk can become confusing when placed in front of a child.

A difficulty that looks like “poor comprehension” may turn out to have started several steps earlier.

And a child who appears unable to understand a passage may sometimes understand much more once the barrier to entering it has been reduced.

We do not yet have all the answers.

But one question is becoming increasingly important to how we build:

Before we ask whether a child understands what they have read, we should first make sure they have actually been able to read their way into it.

That is where reading begins.