Keigo-Yuta
Japanese professional Go players Yamashita Keigo and Iyama Yuta
Credit: 2014 European Go Congress

When it comes to making risky decisions, even the experts listen to the crowd, concludes a study published this month in the journal “Evolution and Human Behavior.” Studying cultural change with techniques normally used for the spread of viruses and genes, the research has implications for high-stakes decision-making in medicine, politics and finance.

Social scientists have long been interested in how social influence affects decision-making, but it's difficult to quantify outside of controlled laboratory conditions. A team of anthropologists, led by UNM postdoctoral researcher Bret Beheim, studied the evolution of strategies within the East Asian board game of Go. 

Bret Beheim
Bret Beheim

"What Go allows us to do is track strategies over decades, even centuries, as they pass from person to person," Beheim said. "The social circumstances at the moment a pro uses a strategy could tell us something about why they decided to use it."  

The research team analyzed a digital archive of tens of thousands of professional Go games from China, Japan and Korea since the 1950s. The team posed a simple question to the database: are the strategic choices of professionals better- predicted by their strategies in previous games, or by what's popular among other players?

"Top professionals are still better at Go than the fastest supercomputers, so we thought they might be insensitive to fads in opening strategy," said Beheim. "They're actually well-predicted by it, about as much as their own past behavior."

This makes sense, though, because new innovations are appearing constantly inside the game and one’s information quickly becomes out of date. Even top professionals are never quite sure if a move that worked last time will work again.  

There's also quantifiable differences across nations. "Professionals in South Korea are better predicted by what's popular on average. It was the opposite with Japanese players though. They are better predicted by their own past behavior, not larger population trends," Beheim said.

The team suspects that this might result from structural differences in social networks in each national league, or differences in institutions and cultural norms across East Asia.  

Beheim became interested in studying the games as part of the Cultural Evolution Lab at the University of California Davis, which studies person-to-person transmission of ideas and technology. On learning of the long history of Go, he began thinking about using game records to find real-world evidence for the theoretical models of social learning developed in the lab. He was able to get access to a large archive of professional matches and used that as a basis for the work.

The team is now planning to compare these results against European and American professional chess databases.