Wolfram's Facebook Big Data Research Is Rich With Insights

rohnjaymiller
Rohn Jay Miller Director of Digital Strategy, Hanley Wood Marketing

Posted on April 26th 2013

Wolfram's Facebook Big Data Research Is Rich With Insights

Stephen Wolfram doesn't shy away from big challenges.

At the age of 21 he was one of the first winners of the MacArthur "Genius" grants for his work on the Theory of Strong Interaction in particle physics with Geoffrey Fox of the University of Indiana. Then he invented the field of complex systems study when he founded the Center For Complex Systems at the University of Illinois in 1986. He built Mathmatica, a software program for solving algebra problems, which is still widely in use today.

But he's probably best known in popular culture for building Wolfram! Alpha, the natural language processing (NLP) search engine he released in 2007. Since then Wolfram! Alpha has been built into Apple's Siri product and Microsoft's Bing search engine.

So it's not surprising that he would be compelled to dive into the ocean of data that Facebook produces. Last year he announced Personal Analytics For Facebook which allows Facebook users to comb through dozens of fascinating graphs and statistics about their individual accounts.

These include ages, locations, and Facebook activities of all friends. In return for providing these personal reports Wolfram asks for access to your data for purposes of big data analysis. (The site terms are clear that your data will be used anonymously)

This week Wolfram released a massive report of what all this Facebook data says about us. (The report can be found on his blog here: http://blog.stephenwolfram.com/2013/04/data-science-of-the-facebook-world/ 

First, of course, there's the caveat that the data are only from the million or so Facebook users who have signed up for Personal Analytics For Facebook. But comparing Wolfram's data with what Facebook has told us shows that his data are closely in line with the general population.

What are the big "a-ha's?"

There's a huge number of people on Facebook who have almost no friends at all. The implication is that someone reaches out to them to be a friend and that was how their accounts were established.

Oh, and there are a big number of teen-agers (and perhaps younger kids) on Facebook and they tend to play fast and loose with facts--like their age and their relationship status. (Note to self: check my daughter's Facebook pages when I get done with this post.)

The average number of friends we have is 342--and growing fast. Wolfram speculates that Facebook's recommendation engine, which cuts out at 200 people, helps push up the average number of friends.

The report has dozens and dozens of charts and word clusters about various demographics and behaviors. But one of the most fascinating analysis Wolfram did with the Facebook data was to model the "clusters" of friends each user has.

For example, I have four major clusters of friends. First are colleagues in New York City from my agency days around 2005 when Facebook was just becoming a major network--there was no one else we knew on Facebook, so we connected with each other. Second are friends in Los Angeles and San Francisco from the 1990s when I was in graduate school at USC and then working in the Bay Area. The third are friends in Minneapolis, where I live now. And the fourth are family members, scattered across the country. So my "cluster" model would look like this: big data

 

It's comprised of the four clusters of people I know. As you can see from the lines my clusters have little "crossover," that is, my friends in New York don't know my friends in California because I'm all they have in common. Wolfram reduced each of the million+ Facebook users he studied into these simple cluster forms and found that the cluster form with the largest number of users was a triangle, followed by two nodes, then three nodes in a straight line (a person with three nodes with little crossover. ) And then the four node cluster like mine. Here's the distribution of these cluster models:   Facebook research

(credit: Wolfram! Alpha)

The report is a trove of data sets that allow broad comparisons not only of Facebook behaviors but what we tell Facebook about ourselves. For example, the older women get the more likely they are to report their status as "single," but the same isn't true for men. The cluster models with the most notes--four--are more likely to be from older people who are active on Facebook. That makes intuitive sense because as we grow older we're more likely to move (as I did) or otherwise build relationships in a new social cluster. Wolfram also looked at the activities we're interested in that we list on Facebook. It's no surprise to me that men vastly outnumber women who like "sports." But it's also surprising that men like movies, fitness food and drink and travel more than women: social network

(credit: Wolfram! Alpha)

Women are more interested than men in friends and family, special occasions, and quotes and life philosophy. Who knew?

There's hundreds of other such nuggets made available this week. If you're interested you can sign up for Personal Analytics For Facebook here.

I can't wait for Wolfram to apply the full power of his natural language processing technology for this data so I can find out how many middle aged left-handed Van Morrison loving lousy male golfers there are on Facebook.

Just like me.

rohnjaymiller

Rohn Jay Miller

Director of Digital Strategy, Hanley Wood Marketing

I'm Director of Digital Strategy for Hanley Wood Marketing in Minneapolis.  We've been in the content strategy and branded content business since 1984.  We deliver across content strategy and branded content. I speak in public, often by invitation. I can be reached at rohnjay@rohnjaymiller.com

See Full Profile >