Teach a computer friend to count – and improve your maths skills while you’re at it!

18 November 2011

Agneta Gulz

Those who have been given the task of teaching someone something learn a lot themselves in the process. Many people know this from experience and it is also well substantiated in research. But the pupil does not actually have to be a person – it could just as well be a little character on the computer. For pupils who have difficulties with maths, a character like Eli could be their knight in shining armour.

Eli is a character in a maths game being tested by cognitive science researchers from Lund University on pupils in years 1–8. By teaching Eli and the other computer characters, the children gained a better understanding of maths.

“Our research shows that it is a very effective learning tool. This applies in particular to the weakest pupils – those who need help most”, says researcher Agneta Gulz.

Cognitive scientists in Lund have tested the maths game in years 1–8. By teaching Eli and the other computer characters, the schoolchildren improved their own understanding of maths.The aim of the maths game is to compete against the computer and receive points for cards that are moved in different ways. The exercises are about the decimal system, which is not always easy to grasp.

“Why does it become two figures, the units and the tens, when you add 9 and 3? Why do you have to ‘borrow’ when you want to do 12 minus 4? These things are obvious to most of us, but for some children it’s not easy”, says Agneta Gulz.

Children who have difficulties with maths may with difficulty be able to learn to count with pen and paper, but cannot do sums in their heads. They also have difficulty putting numbers in size order and feeling comfortable with maths.

Such children can learn a lot from someone like Eli. He is one of the small characters known as ‘teachable agents’.

In Agneta Gulz’s example, ten-year-old Annika is given the task of teaching Eli to do the maths game. First Annika plays on her own, then Eli gets to ‘watch’ and ask why Annika does what she does. In the next step Eli plays under Annika’s guidance, and finally Eli gets to try playing on his own or against a classmate’s computer character.

Afterwards the ten-year-old and her computer character chat about the results. If he has done well, Eli is pleased, and if he has not done very well, Eli wants to try again.

Agneta Gulz sees a number of reasons why the method works so well, especially for the weaker pupils. One is that the children are highly motivated; they think computer games are fun and they want to help their little character pass the challenges. Another is that they receive continual feedback, because Eli and the other characters tell them if they have not understood an explanation. A third reason concerns the children’s self-confidence. If they do well, their self-confidence is boosted, and if they do badly, it is not as bad as getting a normal maths exercise wrong.

“We know that effective learning requires you to dare to try again and again, with the risk of failure. But some children dislike getting things wrong so strongly that they easily lose heart and give up. This doesn’t happen in the game, because they are encouraged by their stubborn little pupil”, explains Agneta Gulz.

It was Stanford in the US that first began researching this type of computer-generated character as a learning aid. The characters have AI, artificial intelligence, which means they learn from their experiences. For today’s computer game-playing children, these characters become natural little friends whom they want to help compete against the computer or a classmate’s character.

Stanford and Lund are now at the international forefront in this field. For those who conduct research on learning, learning games with ‘teachable agents’ provide masses of valuable material. All decisions and all ‘chats’ are saved on the computer and can be analysed afterwards.

“It is much better than hanging over a child’s shoulder to see how they solve a maths problem”, says Agneta Gulz.

The teachable agents don’t have to look human. Stanford has made a science quiz with a type of troll character. However, it seems to be best to make them look androgynous, without any clear male or female characteristics. The reason, unfortunately, is that female characters often get nasty comments such as “what an idiot – typical girl!” For this reason, the agents have neutral names like Eli or Tobi and equally neutral appearances.

Almost all learning games are about science or maths. Agneta Gulz and her colleagues are therefore excited about their upcoming project – developing a Swedish game about history. In the game, the agents will have a certain amount of knowledge from the beginning, but only of life in the 21st century.

“Then the teachable agents will ask the same sort of questions as a real pupil who doesn’t know a lot about, for example, the 16th century. How did people live? Why did they have animals indoors? Did they not even have bicycles in those days? and so on”, says Agneta Gulz.

She hopes that the history game could be of practical use to teachers in a few years’ time. The maths game, which has been tested in a research collaboration with University West, among others, is now being used in a number of schools in Gothenburg and Trollhättan-Uddevalla. It is something in between a research tool and a teaching aid, but the history game, if it is successful, will be developed into a full teaching aid.

Quotes from pupils:

Hannes: They didn’t do very well; we need to teach them better!
Ellen: My character asks so many questions… but of course, that’s because he wants to find out more...
Sara: Yes! That was the card I thought it would choose.
Moa: Arvid, your character isn’t half as good as mine. You know what, let’s teach them together!

Cognitive science researcher Agneta Gulz has seen good results among the pupils who have tried the maths game.

- Ingela Björck