In a few short decades, too much information has become a fact of life.  We are all drowning in it.  How we sift, filter and navigate the massive sea of information has emerged as a profound personal, political and cultural issue.

This past year, Eli Pariser, in his book The Filter Bubble, picked up on earlier messages propagated by Cass Sunstein and others that the personalization engines that, for many of us, serve as tacit and largely invisible information filters on almost everything we search, see and buy online are having a powerful impact on our cultural and political lives.  By showing us information they think we are interested in, based on our previous search, browsing and purchasing history, these filters could sequester us into informational ghettos.  Another concern is that the sheer volume of informational choices will overwhelm.  In response, people will have the tendency to restrict themselves to news and information that supports their existing views.  In both cases, the result is an echo chamber where we see and hear information much like what we have already seen and heard.  Gone are the friction and debate so critical to a vibrant democracy.

Two new research papers dig deeply into this debate, seeking to determine exactly how large the real impact of filtering and information navigation habits are on our online experience.  The results are somewhat surprising.In “Ideological Segregation Online and Offline” (Quarterly Journal of Economics, November 2011), authors Gentzkow and Shapiro compare browsing behavior with offline media consumption to determine how ideologically segregated news consumption online is compared to offline and face to face interactions.    

What they find is that news consumption online is far from segregated.  In fact, online consumers with largely conservative or liberal exposure are quite rare.   “The average conservative’s exposure is  . . . similar to a person who gets all her news from usatoday.com. The average liberal’s exposure is  . . . similar to a person who gets all her news from cnn.com.”  Why is this?  The authors find that most people who consume news on the Internet do so from a handful of fairly centrist sites and tend to consume from more than one place.  Jumping from place to place is easy online, and people do it readily.  In addition, the trend over time has increasingly been a concentration of consumers in a handful of high traffic sites such as Yahoo.news, cnn.com, usatoday.com and msnbc.com.

The authors also find that those consumers who tend to visit sites with more extreme points of view – like rushlimbaugh.com – are also more likely to have visited nytimes.com.   Similarly, liberals who spend time at moveon.org are also more likely to have visited foxnews.com.  Why?  These are news junkies who seem to want opinionated information from a wide variety of sources.   “Their omnivorousness outweighs their ideological extremity, preventing their overall news diet from becoming too skewed."

In fact, in the range of ways people consume news and information, ranging from broadcast news to face to face political discussants, the Internet is on the lower end of the segregation spectrum, just above local newspapers (see figure II).  The authors conclude that the Internet is not segregating people into ideologically polarized groups.

A similar question was asked by Fleder, Hosanger and Buja in “Will the Global Village Fracture into Tribes: Recommender Systems and their Effects on Consumers”.  The authors examine a personalization system for buying online music.  They find that even when the system actively suggests a wide variety of choices, in instances where a recommender system is in place, people tend to buy the same songs.  They attribute this to two factors – the volume effect and the taste effect.Volume is pretty simple to understand.  They say that people simply purchase more songs when a recommender system is active and therefore the chances of them buying similar songs goes up.   The taste effect is a bit more subtle.  The authors observe that after recommendation, if people are buying more songs, then they are also buying a more similar mix of songs, which they call the taste effect.     They did not find any evidence that the larger network with the recommender system was fragmenting into smaller sub-networks or groups at either the individual or the population level.  The authors conclude that recommender systems tend to create “increased commonality”.

Both of these studies raise some interesting questions.  When consumers are faced with a number of new choices, either recommended to them or through browsing behavior, will they tend to naturally congregate toward the middle?  The first study shows that most news consumers tend to like centrist news sites.  Is there such as thing as a “centrist” song, or a set of song choices that are somehow pleasing to the average listener and therefore chosen more often?  In other words, is it possible that there is in fact a kind of human and social architecture that infuses the Internet and guides us towards the midpoint, regardless of the massive amounts of content available at any given moment?  Pattie Maes, author of one of the first recommender systems, alludes to this when she says that we don’t really want a hyper personalized world.  She said, “You don’t want to see a movie just because you think it’s going to be good.  It’s also because everyone is talking about it, and you want to be able to talk about it too.”  Perhaps our need to be participate in the global tribe will, in the end, mitigate the potential of technology to create individual digital islands that isolate more than unify.