How We Teach AI

At ChildFirst, we do not teach technology for technology’s sake. We teach coding and AI because of what they can develop in a child — logical thinking, problem solving, creativity, perseverance, independence and the ability to use increasingly powerful technology intelligently.

Our approach follows a simple progression:

We start with coding.

Coding develops logical thinking and problem solving.

Logical thinking and problem solving develop the child.

We can see some tangible results.

But the deeper outcomes are intangible.

Then AI gives children new ways to extend their thinking.

AI can also personalise their learning.

But the child and teacher remain in control.

1. Coding Concepts Without a Screen

Our youngest children begin learning to programme without a screen. Their first programme is made with physical cards.

Children lay out cards with instructions such as forward, left, right and stop. Then they — or one of their friends — become the robot. Using their own bodies, they execute the sequence of instructions exactly as it has been programmed.

This gives children a physical understanding of what a programme is: a sequence of instructions that someone — or something — can execute.

The next step is to scan these cards with a scanner so that a real robot can execute the same sequence of instructions.

All of this happens without the use of a screen.

The children are already learning to sequence instructions, predict what will happen, test their programme, discover errors and refine it — before they move on to block programming, where they drag blocks onto a workspace and sequence them to create instructions for the robot.

2. From Coding to Logical Thinking and Problem Solving

Coding is a good way to teach about logical thinking and problem solving.

In the beginning, our children often solve coding problems through trial and error. They make a sequence of instructions, see what happens, discover that it did not work, and try something different.

As the projects become more complex, however, trial and error alone is no longer enough.

Take Robot Dance. Children need to coordinate a group of robots to perform a synchronised dance with the music, while making sure the robots do not bump into one another. Suddenly, there are many things to think about at the same time: the sequence of movements, timing, spacing, repetition and the interaction between different robots.

This is where they begin to encounter Computational Thinking — a systematic way of solving problems logically, used by computer scientists.

Through projects such as Robot Dance, children encounter the four fundamental ideas of Computational Thinking: decomposition, pattern recognition, abstraction and algorithms.

Decomposition means breaking a complex problem into smaller, more manageable parts. Pattern recognition means noticing repeated structures or movements. Abstraction means identifying the essential information and ignoring what is not relevant. Algorithms mean organising a sequence of instructions that will reliably produce the desired result.

The children are not learning these ideas simply by memorising four definitions. They discover why these ways of thinking are useful because they are trying to solve a real problem and make their robots do something they have imagined.

And when their first attempt fails, they have to think again. They test, identify what went wrong, change their instructions and try again.

This is also where coding becomes a powerful opportunity to develop a growth mindset.

One of our parents described it beautifully. When she asked her son what he would do if he could not solve a problem, he replied: “Then I’ll learn from the mistakes.”

What impressed the parent was not simply what he knew about coding. It was the attitude behind the answer — a willingness to try.

Psychologist Carol Dweck describes a growth mindset as the belief that abilities can develop through effort, persistence and learning from mistakes. The goal is not to make children think they must succeed immediately, but to help them discover that difficulty is part of learning.

“I can’t do this” can become “I can’t do this yet.”

3. The Tangible Outcomes of Coding

This approach has produced tangible results.

ChildFirst children have been three-time National Robotics Champions.

But the championship results are not the point by themselves. They are evidence of what children can accomplish when they have repeated opportunities to build, test, solve problems, learn from failure and persevere.

The trophy is something we can see. The capabilities developed along the way are much more important.

Don’t underestimate what your child can do before P1.

4. The Most Important Outcomes Are Intangible

The deeper value of coding is what happens to the child.

The real test is not whether a child can remember what a sensor is or programme a particular robot. It is whether the child begins to use what they have learned to understand and solve problems in the world around them.

One ChildFirst parent saw this unexpectedly in everyday life.

His six-year-old son had learned about sensors, programming logic, “if and then” conditions and loops through robotics. One day, while they were in the car, his father started driving without putting on his seatbelt.

His son immediately reminded him to put it on.

His father asked: “How did the car know?”

The child reasoned that the car had different sensors: one to detect that someone was sitting in the seat, another to detect whether the seatbelt had been fastened, and another to detect that the car was moving. He then connected these conditions to the alarm.

The child was not answering a robotics question. He was using what he had learned in robotics to explain a real-world system in a completely different context.

This is the kind of learning we value most. It is transfer of learning — when knowledge and skills acquired in one context are applied to solve a problem in another context, ideally a real-world problem.

Coding and robotics can also open doors later. Children who develop a strong interest and capability may go on to participate in elite coding and robotics clubs in primary school. Their projects and achievements may also contribute to a portfolio that supports future Direct School Admission (DSA) applications.

These are possible downstream benefits. They are not the reason we teach coding.

The reason is to help children become more capable logical thinkers and problem solvers.

5. AI to Augment Human Thinking

Coding teaches children how to give machines instructions. AI introduces a different possibility: children can increasingly work with technology that can generate, suggest, explain and help execute ideas.

We want children to learn how to leverage that capability without handing over their thinking.

A child may have an idea for a story but lack the technical ability to turn it into a finished story, illustration or animation. AI can help bridge that gap.

The child provides the idea. AI can help with some of the execution. The child evaluates the result, changes it, improves it and decides what happens next.

The child remains the creator.

AI can generate. The child should imagine.

AI can suggest. The child should decide.

AI can help execute. The child should create.

The Great SG60 Ice Kachang Disaster! is one example of children using AI as a creative tool. The important achievement is not that AI produced a story. The important achievement is that young children contributed their own ideas and creative direction and learned to work with the technology to turn those ideas into something they could share.

This is AI as a capability amplifier — not a replacement for the child’s thinking.

6. AI as a Revision Tutor

AI can also personalise learning in ways that are difficult for a teacher to do for every child.

One of our applications is AI Chinese.

The Chinese itself is taught by teachers. But once a child has learned a word, recognising it today does not mean the child will still remember it weeks or months later.

Our AI Chinese system acts as a personalised revision tutor. It identifies which words a child has retained and which need more reinforcement. Words that need review can return more frequently through games and recall activities, while words that have been mastered require less repetition.

Every child therefore receives a different revision experience based on their own learning history.

We apply the same principle to AI Math, using AI to support more personalised practice based on what an individual child has already demonstrated, rather than giving every child exactly the same sequence of questions.

This is where AI can do something particularly useful in education: not replacing the teacher, but helping personalise learning for each child.

7. The Human Remains in the Driver’s Seat

The more capable AI becomes, the more important human judgement becomes.

At ChildFirst, AI is a tool. It is not the teacher, and it is not the child.

The child remains in the driver’s seat. The child decides what to ask, what to create, what to accept, what to reject and what to do next.

The teacher remains central too. Teachers understand the child, notice things that technology may miss, ask questions, provide encouragement, make judgements and build the human relationships through which children learn.

AI can generate. It can suggest. It can personalise. It can help execute.

But the human must decide.

Our goal is therefore not simply to make children good at using AI. It is to help them become capable human beings who can use increasingly intelligent technology to extend their own capabilities — while retaining the judgement, creativity, curiosity and agency to decide how that technology should be used.