Tweets Like Us

What do our Google queries and Tweets say about us? The short answer: We're a pretty divided lot in cyberspace -- but then, we kind of already knew that, right? [We're divided offline, too.]

What's new is that we can now tell what unites us, just from visualizing the knowledge we seek on Google and the hashtags we frequent on Twitter. A new data visualization tool called Web Seer, developed by two former IBM colleagues, provides some insights into what it is about the echo-chambers of today's identity politics that keeps us all arguing so much to begin with.

Skeptical? Let's have a look. Go to "Google Suggest" on your browser and click on "web search." Type in the word "why." Thanks to an "auto-suggest" feature that this and many other search engines now use, just typing in the word "why" produces a list of suggested, presumably popular completions. Today's list? "Why...did I get married?" is the most popular question, followed by, "Why is the sky blue?" and "Why do dogs eat grass?" ["Why don't you love me?" also ranks right up there with "Why did I leave Astoria?" -- a question that people who live in the New York City borough of Queens might especially appreciate.]

Amusing? Yes, and according to a new data visualization tool called Web Seer from Flowing Media, a new startup cofounded by Fernanda Viegas and Martin Wattenberg, searches like these also give the Web-curious "a little peek into our collective souls," says Viegas. "Exploring this Web Oracle can be quite revealing of society's fears, curiosities and prejudices."

Let's type in a new question, "Why doesn't he ... ?" When I did this, I got "call me" and "like me" and "ask me out." Sure enough, typing in the question, "Why doesn't she ... ?" surfaces similar completions. The differences, though, are the most telling in this case; Web Seer's visualization of the Google Suggest data spots the male-female divide instantly. Men tended to complete uniquely with "call me back?" and "just leave?" and "like me anymore?" while women appeared to be more interested in why he didn't "text back" and "want a relationship."

Viegas says deeper gender divisions become evident when it comes to family issues. The question "Is my daughter ... ?" is most likely to generate "pregnant" and "a virgin" and "gifted" and "austistic." Type in the question, "Is my son ... ?" and you'll likely get "a homosexual" and "on drugs." Says Wattenberg, "If you play around with this for a while, you start to see a portrait of people's anxieties and, I think, ultimately, a very clear gender division in society."

When Web Seer is applied to party politics, the divisions seem even more troubling. "Are Republicans ... richer than Democrats?" and "evil?" and are Democrats "socialists?" and "communists?" Viegas says that people in the analysis seem"confused, very much so, about party differences." But the areas of agreement, in this case, were just as revealing. According to Web Seer, people who typed questions into "Google Suggest" about either party asked if both Republicans and Democrats "are retarded" and "are morons" and "are destroying America" -- in other words, most who query "Google Suggest" about U.S. politics are similarly, deeply skeptical about the effectiveness and integrity of either party.

Okay, so now let's check out Twitter. Are we divided there, too? To find out, Viegas and Wattenberg analyzed Twitter trending topics over the 2010 Memorial Day weekend; after analyzing the photographs of those tweeting in some 10 Twitter trending conversations, ranging in topic from #wordsbeforedeath to #oilspill, the pair discovered that non-whites and whites were mostly equally involved in some conversations -- but not others. In this case, whites and non-whites were about equally involved in a conversation about #wordsbeforedeath -- but #cookout tweets were being written predominantly by non-whites and #oilspill tweets were being written, overwhelmingly, by whites.

The point here? "It's not to think, wow, there's racial segregation on the Web. We sort of knew that; the Web is a reflection of real life," said Viegas. "The point here is that this level of segregation is just one click away from the Twitter home page and it's happening in the trending topics. The fact that you can look at trending topics and be immersed in conversations that are so separate from one another is something to keep in mind."

Beyond simply insightful, these kinds of real-time "focus group" data visualizations are starting to be used by political groups, as well as by nonprofit causes seeking new supporters and social enterprises in search of some target markets or information about potential trouble spots beyond those which may seem obvious.

Got any further insights to share? Let us hear from you.

Here are Viegas and Wattenberg at the 2010 Personal Democracy Forum in June:

-- Marcia Stepanek

(Illustration, top: istock.com)