I got 500 friends on Facebook
A lot of bloggers are picking up my piece in today’s issue of The Economist on the possible bio-sociological conclusions to be drawn from Facebook data. A few examples are here, here, here, here and here.
The point of the piece: to add a tiny bit to the research debates about human group size.
Just a few words on the backstory of this article: I got the idea in December while chatting to Sheryl Sandberg, the COO (“chief operating officer”) of Facebook. We were just shooting the breeze when I thought of the Dunbar Number, one of my favorite talking points, and a conceit that I’ve used before. Robin Dunbar, an anthropologist, hypothesized that primates form groups to the extent that their brains can compute the many relationships among group members. I think what I like about it is the originality of extrapolating from the ratio of neocortex size to group size among other primates to Homo Sapiens.
Anyway, we all went off over the holidays, and when we came back from the break, Larry Yu at Facebook put me in touch with Cameron Marlow. He had got some complicated PhD title that amounted to, in his words, “computational sociology” and is now the “in-house sociologist” at Facebook. (Since Mr Crotchety and I think alike–great minds?–he serendipitously emailed me an article about the rise of computational social scientists that same week. If you don’t know Mr Crotchety, you haven’t been reading The Hannibal Blog enough.)
What I wanted from Facebook was numbers that might advance the debate on human group size. It was really difficult. Marlow and his team came back with one set of charts that I could not decipher without help. “Simpler“, I said, which will not surprise regular readers. Eventually, they produced a chart that I thought was simple enough.
Here it is: economist-median-network-size (Incomprehensibly, WordPress does not allow me to embed an PDF chart into my blog, so please click through.)
I like it. It shows three diverging lines: the blue for the number of people we track passively (by clicking on their profiles or status updates, say); the green for the number of people to whom we respond (by commenting on a status update, say) but who don’t respond back; and the red for the number of people with whom we communicate two-way (by chatting, emailing, exchanging wall posts, etc).
The conclusion: the more active or intimate the network, the smaller and more stable, no matter how many “friends” you have on Facebook. I wrote my piece around that chart.
To my surprise, my editor then took an extreme interest in gender differences. So I went back to Facebook and had Marlow produce another chart, this one: active-relationship-size-by-gender
It was still simple, but broke out the same information by gender. Yes, just as you thought, women are more social than men.
To my surprise, the editor didn’t go with any of the charts. What can you do? So I went back to Facebook to get actual numbers. (It was like getting blood out of a stone at this point.) And so we ended up putting those numbers in the text.
Long story short, the piece has no chart but it still gets the point across: No matter what technology you use, we still seem to interact with the same small number of people as ever before. Hence the rubric (subtitle):
Even online, the neocortex is the limit
Anyway, if you’re interested look at the charts and see if that gives you more ideas.
Hannibal and Me: life lessons from history 
What would Marlow say about a woman who read a piece in an economist’s blog, warning about privacy issues on social networking sites, and then, despite her 450 friends (former students) did the unthinkable?
Her final status report said… Cheri is…saying good-bye to Facebook. Bye!
Does Marlow track the relief factor that occurs in human brains?
What a fine title and subtitle, reflecting the critical importance of evolution and brain structure in all these new media social issues.
I suppose nature may have optimized us for 150 relationships but I think we’re in the process of redefining “friend”. Initially this is taking the form of very superficial forms of relationships via social networking but I think over the years we’ll find that people use their online contacts to refine and expand their close relationships, and as that happens the Dunbar number will fall.
Surely our neocortex can handle more than 150 “good” friends?
Courtesy of your Economist article, I was able to add a word, of which I hadn’t before heard, “maniple”, to my everyday vocabulary.
I figured you were behind that article, but I disagree!
The average number of friends statistic is misleading in it’s correlation to Dunbar’s number. A different study in <a href=”http://buzzcanuck.typepad.com/agentwildfire/2007/10/facebook-averag.html” Toronto found the average Facebook user has 164 friends, more than Dunbar’s number. I would argue that this more accurately represents social networks. When you take the entire Facebook population, as in your statistics, new users and users who do not have online friends skew the data downward. For those who have most of their real world network online and have been developing contacts for years, it is not uncommon to see people with 500, 600, even more than 1,000 contacts on the network.
This does not prove Dunbar wrong, but it does blur the lines. In my group of friends, the term fraquaintance has become popular. A fraquaintance is more than an acquaintance, but mutual contact is not often enough to warrant the term ‘friend’. Without online networks, fracquaintances would certainly fall back to regular acquaintances, or even fade out of the network entirely. This is interesting, because it suggests that computers may be actually changing Dunbar’s number by easing the computational burden on the brain. Facebook enables you to remember an acquaintance’s name, see mutual contacts, know when to congratulate them on their birthday, and most crucially, determine if they are single. The neocortex may be our limit, but who’s to say we can’t do a little outsourcing?
On another note, what does the data say about those with 600-1000 contacts on the network? Based on what we can see the relationship is linear and not seeming to approach an asymptote.
Cheri, you’d make a great case study for an article. But I don’t know its rubric yet.
JoeDuck: “the Dunbar number will fall” …. Interesting. You’re predicting that online communications will deepen relationships and lead to a smaller equilibrium. I’ve got to think that through.
Phillip Phog: Maniples were the basic Roman fighting unit in the Republican era and a main reason why the Romans vanquished, in particular, the rigid Greek phalanxes. The Greeks had one fixed formation, easily broken by ravines, creeks, boulders, corpses…. The Roman maniples–discrete rectangles of men–could maneuver easily to wherever the enemy was weakest and could turn easily to fight in any direction, even more than one direction at a time. Always below the Dunbar Number, hence loyal to the death to one another and working as one organism…
Jonathan: Interesting study, but not what Marlow told me. He told me: Median number of friends = 80, average = 120.
I’m totally with you regarding fracquaintances. This study is obviously using Facebook data as a poor proxy to biological group size (as apes were the proxy for humans for Dunbar). Marlow wasn’t able to give me info on people with 500+ contacts. Maybe there were too few?
I have no doubt that Marlow is right. But those numbers don’t represent a fully developed community. How many contacts did people have on their cell phones in 1996? I’m saying that if we were to take a population that is not affected by these barriers the numbers would reflect a more ‘biological’ limit and average.
Certainly there can’t be too few people with 500+ contacts, unless the n constraint is in the millions.
If only that stone had some more blood!
J: The idea that the computer is easing the computational burden is very clever. I’ve never facebooked (sorry for the new verb). Is there some computational, dare I say, butt sniffing? Is there any computational screening that might relieve one of his own effort in filtering? Or is it just a time saver (because the interesting data is readily available)?
The graphs are a bit dry. I think you should have made a histogram of stacked faces – smiley faces.
Jonathan:
re: Neo-cortex outsourcing..I love this phrase…I suggest we call it the Jonathan Test.
This phrase fairly describes when people don’t think for themselves and just adopt (and sometimes manipulate) the scholarship of others acting as if it is their own? Whether acting as a statistician or teacher, if you can’t do the math or research and reach your own conclusions, then I propose that you aren’t really using your neo-cortex other than to regurgitate the work of a machine or a scholar.
The effective trial lawyer regularly outsources for information to project an image that they really know about something, when they in fact do not. Example. In a medical negligence case alleging a failure to timely diagnose an evolving myocardial infarction (heart attack), it is easy for the lawyer (with a degree in political science) to take on a highly trained the Defense expert cadiologist with help of a little neo-cortex outsourcing:
Lawyer: “Isn’t it true doctor, that the MB fraction of CPK in the patient’s blood work, suggesting damage to the heart muscle, can show up 8 hours after a heart attack?
Doctor: “Why yes counsel”.
Lawyer: “And can’t an EKG in the early stages of an evolving myocardial infarction present as false-negative?
Doctor: “Why yes counsel”.
Lawyer: “So, if the EKG and Blood work can be normal in the early stages of a myocardial infarction, would you agree that it would have been more prudent to admit the patient in this case for observation to test the admitting physician’s differential diagnosis of heartburn versus a heart attack?
Doctor: “Well, ummm, it wouldn’t have hurt anybody to have done so.”
Lawyer: “Thank you doctor.”
Applying the Jonathan Test, the lawyer in this example outsourced for the knowledge to effectively cross examine the witness, without truly knowing the science of the EKG or the nature of enzymes, why they appear in the blood after heart muscle damage, and the like.
The identification of neo-cortex outsourcing, is a good reminder to all of us, me included, that language counts. If the news, politicians, lawyers, and scientists subjected their conclusions to the Jonathan Test before publication, the information would be more reliable and helpful. Scientists and physicians have had an edge of this for years with the process of peer review.
How refreshing would it be? “I’m Mike Smith, I approved this Ad only after it was cleared by the Jonathan Testing authority who found that I really don’t know what the heck I am talking about…”.
And Jonathan, what say you?
Our brain-space is precious!
I like outsourcing. To use an odd example, I did significant outsourcing when book 6 of the Harry Potter series came out. Many wondered who the mysterious character RAB was. Since I had all the books in the illicit .pdf version on my computer, I simply did a ctrl-f search and came up that there was only one possible character with those initials. For the space of a week, I became an ‘expert’ on Harry Potter to my friends and the internet.
But when it comes to anything related to law, I fear if I make any comments I would fail the Jonathan test immediately.
I have just received this curious reader letter (name omitted for privacy):
Dear Editor,
In response to the February 28th edition article, “Primates on Facebook“, I must admit I am a little confused. Why, with the current ongoing recession in our economy, are we funding studies on the social life of men and women on Facebook? Although the article is entertaining, I believe that these doctors could have put their time and research to much better use. Perhaps a theory on how the average person can survive the current economy?
The news is packed with doom and gloom reports of a sinking economy with failing banks and people losing their homes and jobs like car keys. People don’t want to know about the latest studies on how many of their Facebook friends they actually talk to, they already know that. People want answers on when the economy’s future will look bright or when they won’t have to worry about their jobs and foreclosures. I think Dr Dunbar has some studying to do.
Sincerely,
[Name]
Andreas: Greetings from Hong Kong! (I just moved back). Nice blog and this posting addresses exactly the question I was wondering about.
Whether digital media allows us to break through Dunbar’s number or not, it does allow a broader spread of communication. (This interaction between you and I, for example).
Come back to this topic sometimes soon, it is a rich one.
Hope all well.
Tom
Hi Tom,
looks on your blog like you might have much more to say on the subject. I’m intrigued what nut you’re cracking for that media company you’re working with. Let us all know what theses you come up with. I was just dabbling here.
A
Thanks for the great writing. 🙂
There is a theory* that men tend to have more, but shallower relationships, women fewer but deeper—I’ve not been able to decipher/infer this from your chart…but is there data to encourage or discourage this theory?
*For example, quoted within http://www.psy.fsu.edu/~baumeistertice/goodaboutmen.htm
I’m intrigued and reading that longish piece, Keith. Will respond anon.
Okay, I’ve read the whole thing. Again, you’ll find NO data in the limited research I did with Facebook. But the research in the essay you link to is perhaps the best data there is to make the point you make.
It is a very profound and convincing essay, I might add. Touches on all the hot buttons of “gender” relations and goes deeper.
It all stems from this:
“Recent research using DNA analysis answered this question about two years ago. Today’s human population is descended from twice as many women as men. I think this difference is the single most underappreciated fact about gender. To get that kind of difference, you had to have something like, throughout the entire history of the human race, maybe 80% of women but only 40% of men reproduced.”
From that, you get all sorts of differences between the sexes, from large variance around a mean (male) to small variance (female)–which is the subtlety that got Larry Summers in trouble–to large and shallow social networks (armies, corporations, politics, ie men) versus small and deep networks (families, neighborhoods, ie women).
Thanks Andreas for adding your perspective—it’s an interesting and probably controversial topic.
I’m really enjoying your writings on social networking. I think it’s quite a fair warning, and useful insight. We live in an information society and these issues are very important for a better usage, understanding and coherence between our virtual and real self.
Plus, the blog’s value added discussions… such enjoyable comments, really. Cheers from Lisbon.