In the science of learning it is long known that every student is required to study differently depending on their experience, skills and knowledge. But most training approaches teach the same way to all students. The student needs to adapt to the course curricula, instead of the curricula adapting to the student needs. Elearning makes little difference here, where each student goes through the same learning material. And even much celebrated massive open online courses dictate one learning path. This is a major flaw in school education and online trainings. One attempt to solve this challenge is to work with real-time lectures or online discussions, which in my experience rarely work well. Personalized learning through artificial intelligence can be a promising alternative. Imagine each student has an own private teacher that adapts teaching to the needs of the student. That is what future learning platforms promise to do and Squirrel in China is implementing that approach already on scale.
A personalized learning platform that monitors teaching success in real-time
At the recent AI World Summit I heard a fascinating presentation from Squirrel’s CEO Derek Haoyang Li. Squirrel is an impressive online learning platform driven by artificial intelligence. Derek presented how they teach 1 million students using adaptive learning technology in various school subjects. Squirrel uses a 3 level approach for personalized learning:
- A comprehensive assessment of each student’s learning level (knowledge points), depending on how they understand each subject.
- The curriculum is finely sliced in small steps, each with alternative ways to teach each task.
- The progress of learning is measured in real-time, and the learning platform adapts the learning path accordingly.
Let’s take for example mathematics, a subject with clear logic and many formulas. Imagine a kid that does not learn math in school, but entirely through a personalized learning system that is tailored exactly to her needs. A course that is not a strict path from A to Z. Squirrel has sliced the math curricula in tiny small parts of learning goals and exercises (see image). Students learn new concepts in small steps and depending on the results each student takes his own learning trajectory. For example, you understand the sine function either through the formula or through the visualization, and AI adjusts the teaching method to your preference.
This personalized learning technology can potentially track a lot of student behavior to draw conclusions for learning success. How long did the answer require? How often do you watch an explanatory video? How often do you stop studying? All of these metrics give insight to learning success, and Squirrel uses machine learning to analyze and optimize the learning experience for their students. If it works, you can learn much faster, which is more motivating. Everybody knows the joy they experienced once they learned something and the frustration of being stuck. That’s why it is so important to be able to adapt learning material on the fly, that would be impossible to teach the same way in a school class.
Here is the full presentation by Derek Haoyang Li
Artificial intelligence provides better education than teachers?
Squirrel was able to win in the past years “AI teacher vs. Human teacher” competitions. The students understood learning tasks faster using the artificial intelligence driven learning program. Are you surprised? I am not. Because if you understand that each student has a different approach to learning a concept, it is very difficult for a teacher to teach all kids at once that require very different learning methods. I am sure you remember times in school, where you just did not get a concept and classmates around were wondering what is your problem.
I believe subjects such as math will be taught that way to kids in the future. If you check the Math course from the Khan Academy or the courses from Brilliant you can get a glimpse of how great that can work. I am sure many kids learn math that way with more joy than in school. But it won’t work for all subjects and such algorithmic driven learning platforms won’t teach social skills so important in an ever complex world. For example, discussing the different twists a novel takes in the literature class. Or reflecting on different perspectives with fellow students about questions of morality. I believe AI driven education can help for certain subjects such as math, biology, physics, chemistry etc. and free up time that teachers can focus on building relationship between students. If you have witnessed kids playing math games on smart phones and enjoy calculating numbers you see where we are heading.
The down side of personalized learning through artificial intelligence
Squirrel CEO admitted by the end of his presentation that a membership costs a lot of money. He does not say how much courses on Squirrel cost, but it seems to be unaffordable to a large part of society. Although they have 1 million students, by February 2018 the company had only 100,000 paid subscribers. That means only 1 out of 10 kids can afford the course and that’s a big problem.
One reason could be the costs to offer a wide ranging learning journey. Producing learning material to explain each topic from different angles can cost a lot of money. Who could initiate such a project? It won’t be traditional education institution any time soon. The required investment probably comes from companies, which would further would privatize this type of education. The private tutoring market is probably quite lucrative. But then only some part of society can benefit from it.
At this stage we have only scratched on the potential of personalized learning. The Internet had in the last years an impact on many areas, but education is only slowly changing to these new opportunities. But should we be surprised if the front-of-class teaching still dominates schools and universities worldwide? It is ironic that „dumb“ machines soon will teach us in many things, but it is also certain that artificial intelligence driven education has limitation. Without surprise even Squirrel relies on a teacher for support if the machine does not succeed.