Skip to main content
cheating to survive

For an enterprising postgraduate student, there are a couple of paths to becoming a co-author of an internationally respected study. You can pore over dusty textbooks in the library and spend days in a tiny research lab scrutinizing specimens.

Or you can build a computer-simulation game and compete against hundreds of other competitors in a scientific "tourney" - and win.

That's what two math-savvy Queen's University postgrad students did in 2009, when they became co-authors of Why Copy Others? Insights from the Social Learning Strategies Tournament, published this month in the journal Science. The study found that those who copy have better chances at evolutionary survival than innovators do, a finding drawn from their computer-game model, "discountmachine."

Oh, and if jumping ahead a number of steps on the science career ladder doesn't seem sexy enough for you, there was a €10,000 purse attached.

Who

Daniel Cownden, 26, of Victoria, is a PhD candidate in the math and statistics program at Queen's University.

He's doing his thesis on evolutionary game theory. His research looks at whether you should invest in information now that might help you make good decisions later, and vice versa. "[The tournament]seemed exactly in line with the problems," he said.

Timothy Lillicrap, 28, of Kingston, Ont., is a PhD candidate in the neuroscience program at Queen's University.

His research focuses on motor cortex control: what makes a person's arm move. "Some of the math I end up using in my daily research is very applicable."

What

The tournament was designed by Luke Rendell and Kevin Laland of the school of biology at the University of St. Andrews in Scotland.

The two specialize in the field of social learning, an emerging aspect of cultural evolution, which explores how people learn from observing and interacting with others.

The project, funded by the European Union, set out to ask key questions about social learning, which is widespread in nature.

The challenge was to find what worked best in a complex environment - trial and error or social learning - to acquire adaptive behaviour.

Current theory predicts the emergence of mixed strategies. Copying on its own was considered an ineffective strategy. The received wisdom is that if "everyone's copying from someone who's copied from someone who's copied from someone, you get into a kind of Chinese Whisper situation," said Dr. Rendell, lead study author and postdoctoral research fellow.

How

The team invited submissions from all over the world, with 104 teams from 16 different countries taking part. It was proposed that entrants, or "agents," would play different games, where they could use trial and error (called "innovate" in the study) or copying, as conditions changed in each round. The call for submissions was circulated on Queen's University's math and sciences listserv in 2008.

Mr. Cownden was the first to commit, putting his thesis work on hold to tackle the tournament full force. Mr. Lillicrap soon hopped aboard. Mr. Lillicrap and Mr. Cownden built a computer program, spending hours to work out complex mathematical equations - and then play they did, squaring off against each other, experimenting with different combinations until they discovered the right strategy.

The eureka moment

It came during the third tournament, about three or four rounds in. "We were actually playing for dessert," Mr. Lillicrap said. "It was contentious." He constantly beat Mr. Cownden (and ate a lot of rice pudding).

"I had abandoned all innovating and was just copying," Mr. Lillicrap said. They submitted their final program, based on that strategy, and called it the discountmachine, because it employed optimal control. "Often in optimal control, you'll do something called discounting future rewards," he said.

"You could take an action like eating an apple now and you get some reward from it. And maybe also in 100 years I could eat an apple and there might be some reward associated with that, but it's small. Our creature [thought]about whether it should spend some time in accruing a reward right now or whether or not to spend that time looking around to what other people were doing and learning," he explained.

The outcomes

The pair were confident they would end up in the top 10. "Because it's a formal simulation, you can actually write out a complicated function of what you're trying to optimize," Mr. Cownden said.

But the discovery that copying, or mimicking, can result in evolutionary survival came as a surprise to the lead researchers.

"We were expecting someone to come up with a really clever way to say, 'Well in these conditions, you should copy and in these conditions you should learn stuff for yourself,' " Dr. Rendell said. "But the winner just copied all of the time."

Copying works because individuals are inclined to mimic the top performers, Dr. Rendell says. That's why cheaters steal homework from the smart kid in class. By focusing on the good behaviour, they're using a filter, therefore passing on information they've adapted.

What's next?

"How ideas flow through social networks is about as important an issue as you can imagine at this point," Mr. Lillicrap said, alluding to new academic research on Facebook and Twitter. He was in Britain last week exploring postdoctoral options. "Getting published in Science is about the best you can hope for," he said. Mr. Cownden said he's looking to do further research work with the tournament's organizers.

Interact with The Globe