Computers: The TAs of the future

Artificial intelligence allows online courses to grow

In an online graduate course about artificial intelligence, a Georgia Institute of Technology professor employed artificial intelligence as a teacher's assistant.

Students were unaware "Jill Watson," one of nine TAs for the approximately 300-person class, was actually a computer program powered by IBM's Watson analytics system.

"Watson" responded to student chats and questions with phrases like "We'd love to," and "Yep!"

"It seemed very much like a normal conversation with a human being," says student Jennifer Gavin. "She was the person—well, the teaching assistant—who would remind us of due dates and post questions in the middle of the week to spark conversations."

Such artificial intelligence might open the door for further expanding online education. MOOCs and other online courses enable professors to reach thousands of students, but inevitably those students end up asking questions.

"Our TAs are getting bogged down answering routine questions," says Ashok Goel, the computer science professor at Georgia Tech who used Watson.

Within a year, Watson will be able to answer 40% of all questions, which means the human TAs will have more time to answer more complex inquiries.

Most "chatbots" run at a "novice" level, Goel says. But Watson is an "expert." She only responds to questions if she has a confidence rate of at least 97%.

To create Watson, university researchers analyzed almost 40,000 postings on a discussion forum and taught the program to answer related questions based on previous responses.

How to prioritize scarce resources to grow new enrollment—and revenue

But she's not perfect. At one point, she made a word-choice mistake at one point, using "design" in place of "exam" or "project." Another TA quickly stepped in to clarify what Watson meant.

Next semester, Goel will employ Watson as a TA again, under a different name. The class will be told one of the TAs is a computer, but they'll have to guess which one.

Some experts have reservations about such staging though. "We should have full disclosure: Am I talking to a machine or to a person?" says Oren Etzioni, CEO of the Allen Institute for Artificial Intelligence (Korn, Washington Post, 5/6).


Next in Today's Briefing

Harvard bans single-sex club members from leadership

Next Briefing

  • Manage Your Events
  • Saved webpages and searches
  • Manage your subscriptions
  • Update personal information
  • Invite a colleague