Learning to Read
AI really comes into its own when we need customised solutions at scale. If we can identify use cases in which human intervention is simply not practical given the complexity and diversity of outcomes, AI might just be able to offer a way out. Such as, for instance, to provide bespoke education outcomes.
This article was first published in The Mint. You can read the original at this link. If you would like to receive these articles in your inbox every week please consider subscribing by clicking on this link.
It has been over two years since Generative AI first burst on the scene—and I have to say that I have grown increasingly frustrated with it.
I jumped onto the artificial intelligence (AI) bandwagon long before ChatGPT was launched. Two years on, after having tried virtually every new application and implementation, I am yet to find one I can rely on. I thought AI was going to do my research for me, write my articles and generally help me make sense of the world. Instead I find myself double-checking everything it sends—more often than not discarding its suggestions and starting from scratch. We need AI to solve real-world problems that humans cannot. If it cannot, it will be just a passing fad.
Just as I was about to give up on it entirely, I came across an application of AI in the education space that just might restore my faith in its promise.
But before I tell you all about it, I need to first explain the complex process of learning how to read.
Speaking and Reading
Children naturally develop the ability to speak—organically learning how to associate the sounds they hear spoken by the adults around them with the objects and actions that those words are meant to represent. They, however, need to be taught how to read. In other words, they need active instruction on how to associate the sounds they already know with the squiggly symbols (alphabets) that we use to construct written words and sentences.
Studies have shown that the ability of students to make sense of what they are being taught is directly correlated with the speed and accuracy with which they can read. We call such students fluent readers.
Our entire education system is predicated on children learning to read by the 4th grade so that, having have achieved fluency, they can read to learn. In higher classes, they are expected to not only follow what is spoken in class, but also what is written on the blackboard and in text books. To do this, their brains need to be wired so that they can quickly and correctly decode symbols into sounds to better understand advanced concepts. Those for whom this wiring is incomplete or deficient will lag their classmates in comprehension.
One of the main reasons why students are identified as poor learners in our education system is that they are unable to read as quickly as those around them. They struggle to understand what is being taught in class - not because they are academically challenged, but because their brains struggle to comprehend what exactly they are reading. If that is the case, then all it should take to improve their educational outcomes is a new way to help them re-wire their brain so that they can make these missing connections and improve the speed at which they read.
This is easier said than done. Every child takes a different educational path. As a result, the way one child’s brain has been wired will differ from those of students sitting next to her in class.
There is, therefore, no one-size-fits-all solution to this problem. Even the most dedicated teacher will simply be incapable of providing this level of customised education to every student in her class to make sure that those who are lagging behind can catch up with those who are on track.
Customised Learning
This is where AI can play a role. Today, AI is more than capable of understanding spoken conversation. It can, therefore, not only assess the written comprehension levels of different students, but also devise customized remedial pathways that address the unique deficiencies of each and every child.
This is what the government of Tamil Nadu set out to do with its Mozhigal programme, a teacher-supported reading initiative that has leveraged AI to improve the language learning of children in that district. Since March 2023, when it was launched, it has been offered to over 900,000 children in over 6,000 schools in an attempt to improve their reading skills.
So how does it work?
Children who join the programme are initially placed in front of a computer and given a piece of text to read aloud. The AI records this, and, by comparing what they’ve said to the written text they were given, is able to identify the syllables and phonemes that they are having trouble comprehending.
With this information in hand, the AI develops customised practice material—words and sentences specifically designed to meet the unique needs of each individual child. The child then practises by reading the sentence out loud, taking help from the computer for words that she finds a struggle. As she practises, the algorithm dynamically adapts the sessions to adjust them with her progress, constantly working towards increasing her familiarity with words that she is unable to recognise.
As a result, children improve their ability to read in gradual steps, increasing their familiarity and speed with each round. This boosts their confidence and motivates them to try harder and more complex sentences.
Practical Use Cases
In a few short months, children enrolled in the Mozhigal programme have shown remarkable improvements in reading speed. The students who have been through its learning modules are already engaging more actively in class and seem more confident about their course-work. While it is still early days, there is every likelihood that their academic results will also improve.
The real promise of artificial intelligence lies in the practical uses to which this technology can be put. We need many more applications like this.
