Is a virtual classroom the future?

Vedantu offers an interactive online tutoring platform where teachers give tuitions to students all over the country using a real-time virtual learning environment. Co-founder Pulkit Jain explains their process in detail.

On the platform WAVE—Whiteboard Audio Video Environment…

Every child from 6th to 12th takes two hours of live classes every day in our supplement platform that replaces tuitions. Replays are very important to solidify a concept because children don't understand in the first go. There are also post-class assignments and exercises along with tests, both subjective and objective. Children, teachers and parent are all associated with the platform.

This generation loves clicking on the screen, so we've created a platform where you can do just that to answer any to be answered by a team of 3,000 experts all across the country. We control the entire learning journey of a child, from the classroom to 12 in the night. The real success happens when you handhold a child throughout. It's not about making him sit in the class and then take a test. When a child’s performance goes down, you have to talk to him, handhold him, intervene and give corrective content. You may have to ask him to stop doing something and do something else.

Our team of mentors use the complied data and insights to make a list of students every day, who need some sort of intervention. This is our massive differentiation. Technologically the virtual classroom can have 10,000 students at a given time. But the highest we have reached is 700. Right now the standard is 150-200.

Making sense of a huge amount of data

We get a lot of data from our live classes. The content is built in such a way that it is designed to capture interactions from the student. It's not a one way traffic. It's not just a teacher talking to 200 kids sitting in 200 cities. The system and content both are designed in such a way that every child has to take action every five minutes, either in asking a doubt or responding to something or taking a quiz or giving feedback. All these are very important parameters for us and there are more than a 100 of them. Four of our main parameters are: student engagement, effectiveness of content, effectiveness of teacher and learning outcome.

We even measure the tone and sentiment of a teacher and the energy he is infusing inside the class. Our content is very interactive because it's digital and built in-house. That’s why we know exactly what is being taught at what time. We devised a concept called hotspot where the teacher can make any part of the screen clickable. Imagine if the student has to choose between a rectangle and a rhombus for an answer, the teacher will simply make both shapes clickable. Students from 200 cities will click the relevant shape. The teacher gets real- time feedback while we collect the data.

If 80% get a question right, then you can briefly tell the 20% what went wrong. But if 80% get it wrong, you may have to teach the whole concept again. If most teachers get an 80% correct in their live classes while some get only 30%, there might be something wrong with the teacher. A child doing nothing is also a data point as are the thumbs up, thumbs down, fastest finger options and quizzes.

We create leader boards which the children just love but then everyone will not make it to the top 20 of that board. We send personalized messages to children who get something right after they’ve made a few mistakes. That makes a lot of difference to them. As part of the live class network, we don’t just have teachers and students but academic mentors, who are invisible. The teacher can route the doubts of the students directly to the mentors in real time.

We are compiling a nationwide database to understand what are the types of questions students find most difficult? What is the time of the day when they love to learn the most? What is their active attention span? What students love the most? What analogy was the most effective?

Artificial Intelligence, Machine Learning and Natural Language Processing

Educational institutes are soulful boutiques which are very difficult to scale without humans. AI in its strictest sense is creating intelligence which otherwise humans used to have. When we get into the AI zone, we either have to replace humans or augment them. But then again nothing can be built immediately. You have to train systems and that’s where ML steps in and you have to keep up the training process. But remember, we can’t “totally” replace humans in this field.

We experimented with running NLP on our data of doubts to identify the clusters of the doubts inside the class. For example, we claim to be the only body in the world that actually knows what are the exact doubts the students ask when Newton’s Third Law is been taught in terms of doubt clusters. When I see the data of 10 teachers teaching that law, if certain doubt clusters are always getting repeated, my content needs to be made better here. If the unique doubts are different for one particular teacher, he may have an articulation problem and we will train him.

We have patented a lot of our stuff. There are parameters related to face recognition, content analysis, tone sentiment and handwriting on the screen. We can analyse whether a teacher’s handwriting is clumsy or small or whether he writes too fast. Our core patent is our ability to analyse all this data, process it, run AI-ML models on top of it, to analyse student engagement, teaching effectiveness and learning models. So anything that can be measured with a live session is used.

We began with using a lot of open source tools and building our layers of specifications on top of them. We didn't build anything related to AI right from the scratch. Today a lot of AI is available off the shelf. But to train it and fine tune it for your use case, a lot of engineering effort is required. We also use a lot of off the shelf ML tools. But we are training it with our data.

Reducing stress among students

We reduce stress. It starts with our not having a selection bias. Most coaching centres start with a selection test and then form benches which are physically visible, top to bottom. It basically will start the learning journey of a child by tagging the child. But in a virtual environment, there is absolutely no tagging. Every child gets a personal attention and this has a positive effect on them. They all love to air their opinions today. Even if a child is scoring less, he gets his opinion heard.

Second thing is being a virtual class the depth of the physical classroom is suddenly diminished. In the physical classroom, you may have 50-100 kids. Is the teacher is far away or near? Is the teacher looking at you or not? These don’t matter in the virtual classroom as the teacher is right in front of you. That engages children a lot.

Third, in the physical classroom, children may shy away from asking doubts for fear of looking stupid. But in a virtual environment, the percentage of students asking doubts or airing their opinions inside of the class, or engagement, is high at 92-95%. Fourthly, our course is designed that our students can ping their academic mentors at any time. This has created a fun virtual environment in the home in the company of parents.

Finally, we do tonnes of motivational classes. Students are very enthusiastic. After the Kashmir internet blackout, students from there called asking for extra classes because they had missed their regular ones. We have additional courses on reading speed, turbo math, fast calculation etc. We even have small Tik Tok videos which explain a particular concept in a short span of time.

On mobiles and the future…

It varies, but 25-35% students attend the classes on mobile. This could be due to convenience or maybe it’s the only device they have. When I asked a student whether it was difficult to study in the mobile, he replied, “But why? I can even play PUBG on my mobile!” More than 90% of the free classes are taken on mobile and that proves to be a good entry point for the students.

We have a great academic system on paper, but the implementation and execution is a problem. We can solve that. So with an NGO, we went to a small school in a rural area. We put up a TV screen in the classroom and our teachers taught there with a mentor and it worked beautifully for months. That was a successful experiment. We can actually go to the grassroots and get vernacular teachers in the future.

Be it in the classroom or the house, now the best teacher of the country can actually teach any kid of the country. School should be a place to teach child social skills, playfulness, physical health etc. But just the academics of it, we can solve it with our virtual classrooms. We feel that our online method will also work much better for a college student. If we overcome the regulatory issues, then online education can really big. In next two years, online could become a sort of de facto way of learning. This will lead to a massive disruption in the education space. In 10 years we could have online universities giving degrees.