Can Machine Learning outdo teaching?

Four years ago, when I was heading the MOOCs division of my University, we had a bunch of young people, running a start-up on graphics and animation based educational courses, give us a presentation. They explained human embryonic development, through graphics in a short film of fifteen minutes. I was amazed by the effectiveness of dissemination of knowledge through graphics. But a disturbing thought was also lingering in my mind, what I had seen was much exciting than the boring biology lessons, I had received in school, and was delivered in a much lesser time too. Would e-content take over class room teaching? The question lingered in my mind and I asked it to our senior Professor. He shrugged, feigning indifference and said, 'For how many years have you had a Guru Shishya parampara?' He meant that nothing could take over the human element in the learning process, and no technology can replace a 'Guru'. Nothing else could be more reassuring to me. I believed him.

Just a few days ago, this thing called 'Machine Learning', managed to occupy my thoughts once again. Machine Learning is a branch of AI, which empowers computers to pick up patterns, that they were not explicitly programmed to perceive, or simply speaking, computers can pick up patterns that human mind will take long to perceive, just by doing some data analytics. Education Initiatives, an Indian company has developed a code called Mindspark, that is run on 2.5 million answers to a set of 45000 questions, to diagnose common errors that the students tend to commit. Mindspark will infer a thinking pattern that is responsible for a certain error and will suggest remedial exercises. It is basically, getting into the mind of a student, studying his thinking pattern and structure, identifying the possible reasons for it, and suggest a solution.

The students can receive customized solution through this technology. Mathematics and Language can be easily taught to a student in his early years of learning, through this software. Customized exercises can be generated to handle individual problems. Many American companies like Aleks, Knewton and Dreambox Learning, Geekie a South American company, Byju's from India, and 17zuoye from China are providing such solutions. Voice Recognition systems will make the technology much more user friendly. What this technology right now cannot do is that friendly pat on the back that the teacher gives when a student becomes successful with something that he is struggling with, or a 'raise your hands' punishment that one received for uttering something really stupid in the class. From an older generation's point of view, some of the best memories of school life will be eliminated by the technology, but that is an older generation's viewpoint.

Studies have shown that software assisted teaching pedagogues are more successful. Philip Oreopoulos, and Andre Nickow for J-Pal conducted 41 randomised control trials comparing students taught by software assisted methods, to students taught by conventional methods, and found that the software assisted students got better scores. In one such study done by J-Pal in Indian settings by Karthik Mularidharan, Alizendro Ganimian, and Abhijeet singh, it was found that at an Indian after-school, course, students assisted by Mindspark, brought better results in Math and Language, than those who didn't, for a fraction of the cost. Studies have also shown that in softer subjects, Humanities and Liberal Arts, mimicking a teacher is a little difficult.

A teacher cannot provide an individual feedback to a class of students more than 30. That is the reason 30 is an ideal class. The software replaces a teacher in providing individual feedback. But the critical success of software lies in the fact that a student is assisted further in learning at his own pace, without getting bogged down by the performance of others. But sometimes digital displays can create a lot of confusion in the mind of a student. Max Ventilla, a former Googler has done just that, he has created a school called the Altschool in California, where the student is assisted in personalized learning. Teachers save time by not marking or planning lessons and the students learn from each other. However, giving children such attention is not cheap, the cost per student is $27000 per year.

We live in a Googleable world. A child growing up in this world might be tempted not to fill up his mind by information. Would that mean he would be spending more time on thinking? Cognitive scientists argue that human minds are not built to think and thinking hard about things does not come naturally. There is a famous quote by George Barnard Shaw,' I Have Made an International Reputation for Myself by Thinking Once or Twice a Week, because others attempted to think only two or three times a year.' 

Rand report, a report on a study done on school children using high tech personalized learning at 62 schools said that those who used the tech fared better than the children from ordinary schools, particularly those who were at the bottom of the class. Personal Help, argues Mr. Mark Zukerberg in the form of tech solutions is way better. The early adopters prove it. Whether Rand Teachers were highly motivated teachers or not is not proven by the study though.

We live far away from silicon valley, so we are curious about it. Will children of my family be able to perform well in the clog wire of machines, AI and codes in future, I am forced to think. I am uncertain about the answer. But, I was a few days ago telling my brother to get my niece trained in music and art. Information, she will gather with not much difficulty in future. Her academic abilities are explained in far greater detail in her report card these days, anyway.

Will tech make the craft of teaching better? We all remember our favorite teachers by not 'what' they taught in class, but 'how' they taught it in class. Unanimously, we would agree that they were favorites because they gave us personal attention.