Tuesday, July 20, 2010

Robot teachers – endlessly patient, CPD updates in seconds

The article in the New York Times will no doubt drive traditionalists to apoplexy, so there must be something in this report about the rise of robots in schools. Imagine a teacher that is endlessly patient, always teaches the correct skill in the correct way, provides lot of constructive feedback, can receive CPD updates in seconds, never gets ill and costs less than one month's teacher’s salary? That’s the long-term promise.

What makes all this possible are advances in AI, motion tracking and language recognition. We have already seen what Microsoft have done with Natal in terms of speech, gesture and motion recognition. Plonk this into a robot and we’re really getting somewhere. The point in the short term is not to replace teachers but to replace the teaching of SOME TEACHING TASKS.

The focus, for the moment, is on early years education, playing to the ‘cute’ factor. This makes sense. We could see significant advances in early numeracy, literacy and second language skills in schools with software that provides guidance superior to that of many teachers. In addition, they can be updated, wirelessly, in seconds and can even learn as they teach.

The basic premise is sound and was shown convincingly by Nass and Reeves in The Media Equation – we treat technology as humans if it has the right affective behaviours – good timing, being polite, co-operative etc.. This is how computer games work. With the right movement, sounds and behaviour, avatars and robots seem real. Tens of millions use and experience this phenomenon every day. There’s even a touch of this in your ATM, where you’d rather deal with a hole in a wall than a live teller.

Robots don’t get hangovers, don’t take holidays, never discriminate on grounds of gender, race or accent. They’re patient, scalable and consistent. The ideal teacher!

Robots & language learning

The initial trials have been in learning tasks in autism and English as a second language. S Korea sees English as a key skill in terms of growth and its Institute of Science and Technology has developed Engey, to teach English. Their goal is to have an effective robot, that is better than the average teacher, in 3-5 years. This is part of a general push in robotics that sees robots do things humans do in areas such as production, military, health and education. Hundreds of robots are already in S Korea's 8,400 kindergartens and the plan is to have one in every Kindergarten by 2013.

The University of California has been doing studies with a robot called RUBI, teaching Finnish. Initial studies show that the children do as well in tests as children taught by real teachers on specific language tasks. The retention was significantly better after 12 weeks with a reduction in errors of over 25%. Another interesting finding is that the robots need not look like real people, in fact hi-fidelity seems to be a little ‘creepy’ for kids. Although for an amazingly life like robot watch this.

CES 2010 featured a wonderful talking robot that follows you around the house and teaches you languages. It was remarkably sophisticated, with voice recognition, face recognition, picture recognition (show it a picture and it will say the word in your chosen language).l

Robots & autism

In a collaborative Japanese/US research project, children with autism have been shown to respond positively to synchronised behaviour from a robot. This is used to move the child on to other types of social interaction. In the University of Connecticut, a French robot is being used to with autistic children using mimicry to establish trust. Have a look at Beatbot's Keepon robot designed for kids with autism.

Robots and personalised learning

Personalised learning can also be realised through one on one interaction and the robot engaging in conversations and learning from the learner. Work of this kind has been going on at the Georgia Institute of Technology, with a robot called Simon. The improvements in AI and natural language processing have led to results in the robotic world that promise one to one tuition in the future.

Robots & physical tasks

There’s also the teaching of physical tasks, such as setting a table, where Honda Labs have taught older children to complete the task without the aid of teachers. Robots can already complete physical manufacturing tasks way beyond the physical capability, speed and accuracy of a human. We’ve had 25 years of robotic surgery, with robots being used to do surgery at a distance, unmanned surgery and to minimise invasion. In May 2006 the first AI doctor-conducted unassisted robotic surgery on a 34 year old male to correct heart arrhythmia. The results were rated as better than an above-average human surgeon. The machine had a database of 10,000 similar operations, and so, in the words of its designers, was "more than qualified to operate on any patient." The designers believe that robots can replace half of all surgeons within 15 years. In January 2009, the first all-robotic-assisted kidney transplant was performed at in the US by Dr. Stuart Geffner. The same team performed eight more fully robotic-assisted kidney transplants over the next six months.

Conclusion

It is only natural that robots, which have replaced highly skilled tasks in manufacturing, should be considered for teaching. Automating repetitive, difficult and dangerous tasks has always been technology's trump card. If we know one thing about teaching, it’s that it is difficult and demanding, leading to unnatural levels of stress and illness. If we can, at the very least, relieve the pressure on teachers, that is surely a noble aim. In its own way, simple robotic, screen programmes like BBC Bitesize and e-learning have already automated a lot of education and training. Robots promise to personalise this process. Every passing month sees improvements in movement, gesture and language recognition, with the technology appearing in the games world this year by Christmas. I have no doubt that robo-teaching will be common in schools in my lifetime.

2 comments:

Simon Bostock said...

I'm with you. There's no reason to suppose that kids and adults can't learn from advanced (or even primitive) machines.

My son's taught himself to read and write Japanese using toys that his grandmother sends over from Tokyo, for example.

I just posted a startlingly similar post on Auto-education based on Sugata Mitra's work.

He has a great thought experiment on remote presence (and using ROVs) in this PDF:
http://www.academypublisher.com/jetwi/vol1/no1/jetwi01015559.pdf

Neither of us mentioned Augmented Reality either. Which I suspect will change everything in a way almost impossible to overstate - what place training when you can augment pretty much every object on the planet with Performance Support?

vtlau said...

It sounds a lot of hype but what is the difference with like e-learning??