In the 1980s, pop culture experienced a brief love affair with post-camp science fiction. Some results were good, like the Back to the Future trilogy and The Terminator. And some were not, like the comically horrendous Fox sitcom Small Wonder, which featured a robotics engineer who was sheltering his robot daughter from an unfriendly outside world. With nosy neighbors and meddling co-workers wanting a piece of the family innovation, Vickis existence was fraught with peril but her purpose for the family seemed to be to provide a playmate and homework buddy for said engineers 12-year-old son. After all, no one can blast through a fractions worksheet like a robot, right?
The robotics experts at CMU have yet to create such a lifelike robo-friend as Vicki but they may be on pace to create a digital tutoring system that students wont have to protect with contrived situation comedy. This week, the School of Computer Science hosts back-to-back conferences, both aimed at exploring the future of digital tutoring and data mining in education. The Third International Conference on Educational Data Mining and 10th International Conference on Intelligent Tutoring Systems will welcome experts from more than 20 countries to explore new research and further examine the role of robots and computers in the classroom.
These systems can build up a model of the student, of what the student understands and doesnt understand in order to focus instruction where its needed, says CMU Research Professor of Robotics Jack Mostow. These tutors dont just give the complete answer and score it. They get to watch you solve the problem step by step.
With each movement of the mouse or click of the keyboard, data mining systems keep second-to-second records of your path to solving a problem. Every math teacher asks you to show your work, but educators have never been able to map the process from question to answer in real time. Mostow hopes these conferences will further improve this cutting edge technology, to not only help students more effectively but to inform teachers about trouble spots so that every facet of education is more effective. That would truly be a small wonder.
Human tutors are so much more effective than classroom instruction because they can get so much more information about the student, Mostow says. While schools may have test score data over the course of several years, these intelligent tutors capture data over an entire school year but minute by minute or second by second so it is much more detailed. We are working to confront the challenge of extracting useful information out of all that data, finding the patterns that show you how students really learn.
Source: Jack Mostow, CMU Robotics
Writer: John Steele
