Pay attention, the computer is watching

By FRANCIS TILL

We've all seen it and most of us know what it feels like - that glaze that settles on the eyes when a lesson or a lecture turns to gibberish.

Usually, the teacher responsible for the episode of woolgathering doesn't notice and continues droning on - and if you happen to be working through an e-learning module, there's a 100 per cent chance your computer won't realise that you've abandoned ship and gone out to the garden.

That's about to change.

A team of 24 Massey University scientists, led by senior lecturer and computer science PhD Abdolhossein Sarrafzadeh, is in the final stages of developing a program that will transform e-learning by giving computers the equivalent of a good tutor's intuition.

The scientists call it an "affective tutoring system", and Sarrafzadeh says it will give e-learning programs the ability to read the reactions of students as a lesson unfolds, tailoring the presentation to keep interest high.

If that sounds a little like the Machine used to detect replicants in Blade Runner, you've got the idea.



Sarrafzadeh says the new system will be able to keep track of facial expressions, body movements and biological cues such as heart rate to determine whether students are learning.

If they are not, the system will alert the e-learning program, which will change tack in an effort to pull the student back onto the learning curve.

"As it stands now," Sarrafzadeh says, "computer learning experiences are inferior to classroom experiences, which are in turn inferior to tutorial experiences where a student and teacher interact one to one."

The affective learning system will be able to reproduce many of the elements that make tutorial teaching so effective, he says, because it will respond to the student's learning state by modifying the lesson plan.

"We start by building a model of the learner, including the knowledge state and the internal state," he says.

"If we see weaknesses in the knowledge state, we'll be able to predict likely areas of confusion in the lesson, and the internal state model gives us something to measure biological and other information against as it develops."

Sarrafzadeh agrees there are also obvious applications in the security field, although the current system is aimed at learning, rather than interrogation.

The tools used to do this are an "intelligent" mouse that collects information from skin registers much in the manner of a lie detector, and a webcam that keeps track of facial expressions and body language.

The data is now being processed through the computer delivering the lesson - trials have started using maths lessons - but Sarrafzadeh hopes to incorporate all the system's operations in a stand-alone box.

In part, that is because it makes huge demands on the computer's processing power.

But a stand-alone box that works with most computers is also a marketing dream, and Sarrafzadeh expects sales in the millions.

So far the research has been supported by Massey, with the co-operation of Auckland University, AUT, the University of Canterbury and business e-learning company The Learning Curve.

But Sarrafzadeh is looking for outside angels to bring the product to commercial fruition.