Teach Secondary Issue 14.3

THE FUTURE IS NOW Beyond automated busywork, what else can AI do for educators? Patient tutoring and incisive remediation, according to Learno.ai co-founders Madhav Kaushish and Bharat Dhody ... In your view, what would a genuinely productive and useful implementation of AI tools within a school look like in practice? Madhav Kaushish: It’s important to ask howwe can create something which removes a lot of problems that teachers and students face before being able to unleash their creativity. One example would be what we and others are doing to automate different tasks that can be necessary, yet tedious and repetitive, and which don’t necessarily create direct value in terms of educational outcomes. Bharat Dhody: AI enables teachers to individualise learning at a level they never could before. If you want to give 20 students a different tailored assignment each, your first problem is one of diagnosis – choosing which assignment to give each student. And then you have to go and actually create them, which can be extremely time-consuming. To use a simple example, there’s one kid in the class who likes football, and another who’s into cricket. If I’m teaching economics, I could task the ‘football kid’ with telling me howmuch David Beckham sold for and why. ‘The cricket kid’ could meanwhile look into how much Sachin Tendulkar’s sold for and why. If a teacher tried to create an individualised assignment for every one of their students, it could take days. The individualisation possibilities of what AI can do are really exciting. But when it comes to figuring out what a student does and doesn’t know, and then remedying it – that’s where the big changes are now happening. Before, if we wanted to find out whether a kid knows Pythogoras’ theorem, we might give him 10 trigonometry questions, assuming that will tell us, and he gets a number of the answers wrong. Why is that? We might then set a further 10 questions on trigonometry and another 10 on algebra to try and locate where the student’s misunderstanding might lie. Using AI, the student instead gets just one question. If he makes a mistake, then just like a human teacher sitting beside himwould, the AI edtech can query his response. If the student then voices a basic error or misconception, the AI can ask why the student thinks that. Because it’s conversational, the AI can probe, there and then, exactly where the student’s misunderstanding might stem from. Let’s say we’ve identified that the student has issues with understanding algebra. Before, a teacher might look to remedy this by having the student either read a textbook chapter, or watch a pre-recorded video that explains the concept. But there’s a good chance that this algebra video only spends a minute or two of its running time on the student’s misunderstanding. Plus, it’s a one-way explanation – if the student still can’t understand what the textbook or video is telling them, there’s nothing they can do. AI, on the other hand, can extend to history, where students will be expected to write sophisticated essays that present a series of arguments and end with a reasoned conclusion. It’s now “Figuringoutwhat a student does and doesn’t know– that’s where the big changes arenowhappening” create content tailored to the student and address their misunderstanding more forensically. If the student still has difficulties with the AI’s initial explanation, it can try again, this time presenting the information in a different way. That ability to zero in on specific misunderstandings is certainly powerful – could AI perform a similar function in subjects that are more essay-based? BD: If anything, the technology is actually better in areas other thanmaths. The same process could 38 teachwire.net/secondary