How Teaching Assistant Use AI

May 15, 2018 | 1923 Views

Children of the elite used to go to small schools to learn from master teachers. Every student got personal attention, but the education was accessible and affordable. Only by a few with the Industrial Revolution, education reached the muscles accessibility, affordability became big issues, so education had to scale up ready cost of personal attention. Today we have many classes with 100 or more students. Few students get personal attention. We even have massively open online classes in which students learn from pre-recorded video lessons. Some of these classes attract a hundred thousands to, but only 10 % or less completes because of lack of teaching assistants and personal attention.

This raises a fundamental issue for us: can we have personal attention at scale? What, if what if we could give personal attention to every student when and how he or she needed it now? That would create educational revolution; create educational revolution, because learning and teaching would become personal. Goal of AI is to use AI to make education fun. It all began in January 2014, when Georgia Tech started a new educational program. It offers program offers graduate level classes professors record video lessons. Students can access these video lessons through a company called audacity. Professors, teaching assistants and these students engage with each other through online discussion forum to edit by another company called Piazza in 2014. It turned out to be what are the most popular courses in this program about 350 students take this course every semester. By now, about 2,000 students have taken it when we offered fit for the first time we had a surprise. The discussion forum is extremely active. 

The discussion forum acts as the virtual classroom and with students and teachers and teaching assistants engage with each other to conversations, debates, and discussions. It is a real beating heart of the classroom. Each day for the next hundred days, a rough estimate is that would take a teacher working full-time for a full year to answer all of these messages. Therefore, in this summer 2015, I thought that perhaps we could develop an AI teaching assistant that could automatically answer frequently asked questions. All of us teachers know all those of you who raised your hands. All of us know that students ask these same questions again and again so we thought if we could automatically our answer this repetitive questions, then we could free the teaching staff to focus on more creative aspects of the course. However, answering repetitive, questions is not easy. Here are three superficial forms of the same question same question, three forms you and I, as humans immediately get it. These are three simple, functional forms of the same question. How would an AI understand it? How do we teach an AI that this is the same question and therefore the AI must develop similar answers to all three questions. We started working on this automated artificial intelligence teaching, a system that we call Jill Watson and fall of 2015 from IBM. Many people know about it from Piazza. All the data and identified many categories of frequently asked questions. The pace picked up in late fall of 2015. First, all the question-answer pairs were taken from the previous semesters and by now there were thousands and thousands of them and we gave Jill a memory of discussion around Surveyors. Then this memory was organized into many coefficient. Many categories of different kind of questions, then, as a new question, would come in with brain Jill, so that Jill would classify that new question into the appropriate question category. Finally, Jill found an answer from her memory of previous question answers and calculated a confidence value in her answer. This is how she works.


Source: HOB