Educators Can Leverage AI to Build Intelligent Tutoring Systems 

Artificial Intelligence is significantly making its way into the education sector. The technology in education has the potential to perform certain tasks efficiently, transforming the way of learning for students by delivering enhanced capabilities. Recently, a team of researchers at Carnegie Mellon University has developed an AI-powered technique that enables educators to rapidly create intelligent computerized tutoring systems. By showing various ways to solve problems in a topic, a teacher can teach the computer, such as multicolumn addition, and correcting the computer if it responds incorrectly.

The computer system can not only solve the problems as it was taught, but also take a broader view to deciphering all other problems in the topic, compared to a teacher. A Ph.D. student in CMU's Human-Computer Interaction Institute (HCII), Daniel Weitekamp III explained that "A student might learn one way to do a problem and that would be sufficient. But a tutoring system needs to learn every kind of way to solve a problem." It needs to learn how to teach problem-solving, not just how to solve problems.

The advances in AI have tempted great contributions from academia and industry in recent years. In this regard, the deep neural network-based machine learning techniques, enabled by the considerably surged computational power and available data, are increasing and employing to facilitate more intelligent systems. Intelligent tutoring systems are designed to incessantly track student progress, provide next-step suggestions and pick practice problems to assist students to learn new skills through play.

According to Ken Koedinger, professor of human-computer interaction and psychology, when building the first intelligent tutors, he and others programmed production rules by hand, a process that took around 200 hours of development for each hour of tutored instruction. Afterward, they would create a shortcut, wherein they attempted to demonstrate all possible ways of solving a problem. That cut development time to 40 or 50 hours, but for many topics, it is practically impossible to show all possible solution paths for all possible problems, which lessens the shortcut's applicability.

Reportedly, the new computer system may enable a teacher to create a 30-minute lesson in about 30 minutes. Ken termed this a grand vision among developers of intelligent tutors. He said, "The only way to get to the full intelligent tutor up to now has been to write these AI rules. But now the system is writing those rules."

For the Conference on Human Factors in Computing Systems (CHI), which accepted the paper describing the method, the authors demonstrated their method on the topic of multicolumn addition, but the underlying machine learning engine has been shown to work for a variety of subjects, including equation solving, fraction addition, chemistry, English grammar and science experiment environments. The method not only expedites the development of intelligent tutors, but promises to make it possible for teachers, instead of AI programmers, to develop their own computerized lessons.