Sunday, May 15, 2016

Teaching with robots

The potential of big data and learning analytics to radically change our approach to learning and education is moving from the trend reports to practical applications. An article in the Wall Street Journal, Imagine Discovering That Your Teaching Assistant Really Is a Robot, describes how a class at Georgia Tech got prompt and useful feedback in discussion forums from a teaching assistant called Jill Watson and noone realised that she was a robot until the professor revealed the truth at the end of the course. Students post an awful lot of questions and ideas during a course and even if teaching assistants are hired to help with feedback it's hard to keep up. Most student posts require fairly simple responses, often about deadlines, assignment criteria or asking for confirmation that they are on the right track. The motivational effect of getting prompt replies is well documented and today's artificial intelligence can provide this round the clock.

“Our TAs are getting bogged down answering routine questions,” said Mr. Goel, noting that students in the class typically post 10,000 messages a semester.
Mr. Goel estimates that within a year, Ms. Watson will be able to answer 40% of all the students’ questions, freeing the humans to tackle more complex technical or philosophical inquiries such as, “How do you define intelligence?”

Jill had been rigorously programmed by analysing questions and answers in course forums and was programmed only to respond to questions that "she" had a 97% certainty of an appropriate answer. More advanced questions were left to humans. The only hint that Jill wasn't quite like other teaching assistants was that she was so prompt in answering and never seemed to sleep. The students were all positive about the machine responses and many were convinced that Jill was a highly competent PhD student. The robot assistant was convincing though it would be interesting if students would have been equally positive if they had known from the start that they would have a robot facilitator. Of course we prefer human assistance and we all have experience of less advanced and extremely limited chatbots on commercial websites but the alternative for the students in this case could be much less feedback and longer waits for information.

Artificial intelligence is already taking over many mundane and time-consuming tasks in education. Automated feedback on written assignments has also been tested with positive results (see my post from a couple of years ago) and much more advanced applications are in the pipeline as learning analytics matures. Machine translation and speech-text-speech applications are improving and becoming more reliable. Add these elements to the MOOC model and we get scaleable, personalised, collaborative and flexible education where teacher and machine support complement each other This doesn't mean that we'll all learn solely through digital devices and it doesn't replace face-to-face meetings and interaction (arguments I only hear from tech skeptics). It means that we can widen the reach of education and offer alternative paths, integrating education with work and enabling people to learn without uprooting themselves to a university campus (unless you really want to do that). 

No comments:

Post a Comment