As the partisan divides in American society continue to deepen, a team of Stanford researchers has developed an algorithm demonstrating the process behind that polarization — and created Internet-based social systems to counter the trend.
The team, composed of Associate Professor of Management Science and Engineering Ashish Goel M.S. ‘98 Ph.D. ‘99 and doctoral candidates Pranav Dandekar and David Lee, examined how people form opinions and how their viewpoints evolve over time due to outside influences such as social networks.
“When two people look at a blog and as a result their beliefs become more extreme in opposite directions, the blog has a polarizing effect,” Dandekar explained. “We wanted to come up with a model to explain this effect.”
Before developing their own model, the researchers looked at existing models that attempt to describe the formation of public opinion based on homophily, the sociological theory which assumes that “like seeks like.”
However, homophily models predict increasing uniformity of opinion over time, as people seek to minimize disagreement within their networks of friends and relations and as holders of similar opinions tend to aggregate. Those attributes made the model unsuitable to explaining America’s polarization.
“It turns out that you can’t have polarization with this kind of model,” Dandekar said. “When an individual’s opinion converges to an average of those in his or her network, depolarization happens. Instead, polarization is when opinions diverge.”
The team sought to create an alternative model to explain polarization based on biased assimilation, a different psychological concept.
“We wanted to capture this phenomenon that leads to polarization in the way that we define it,” Dandekar said. “The model would finally provide an adequate answer to the question of whether or not people have some biases that they naturally gravitate to.”
According to the theory of biased assimilation, people favor evidence that confirms their beliefs and reject information that goes against their preconceptions, which explains why two people can come to completely different conclusions given identical evidence.
“In seeing mixed evidence, meaning both the pros and cons, of issues like gun control or nuclear proliferation, people usually take in information that affirms their existing viewpoint and reject whatever that goes against it,” Lee said. “At the end, people have more extreme opinions.”
The researchers used their knowledge about polarization to create Internet-based social systems that allow people to identify similarities — instead of differences — between their political views and collaborate on solutions to issues such as the budget crisis.
“The important question is when people have to come up with a budget, it’s difficult for them to do it because of polarized views and disagreements,” Lee said. “We are finding a way to promote consensus in such situations.”
One system developed by the team is a federal budget simulation called Widescope, which allows users to propose their own budgets with detailed annotations explaining their viewpoints. Through Widescope, users can collaborate on budgets and eventually vote on the proposals.
“It’s like a process of elimination. Eventually the best budget proposals will make it to the end through a series of voting procedures,” Dandekar said. “The goal is to come to a consensus and collaborate to find solutions in a democratic manner.”
Lee and Dandekar said that Widescope was inspired by direct democracy systems, in which the entire electorate takes part in making laws and other political decisions. Through Widescope, Americans can simulate playing a larger role in the legislative process — including learning how to come to a consensus on divisive issues — than they might otherwise do.
“In the last few years, there has been an explosion of social networks,” Dandekar said. “The question here is — can we harmonize this engagement in social networking to drive public policy? Can people come together on these Internet-based social platforms to discuss an issue and come to agreement?”
Both Dandekar and Lee agreed that a website like Widescope has the potential to generate relatively accurate representations of the public opinion on a wide range of political issues, from health care reform to federal budgets.
“We are interested in whether or not we can use the Internet to promote consensus and power democracy,” Lee said. “Aggregate solutions from the crowd can be a very effective one. So can we use the wisdom of the majority to make decisions?”