Earlier this fall, Alyssa Milano — known for being on “Who’s the Boss” and, more recently, for being on Twitter — sent out a somewhat surprising tweet to her nearly 1.2 million followers: a link to the Amazon page of a book called Connected: The Surprising Power of Our Social Networks & How They Shape Our Lives.
For a book like Connected, penned by two social scientists and built on longitudinal research and academic inquiry — a book, in other words, that may hope to achieve influence over our thinking, but doesn’t aspire to huge sales numbers — you’d think that a message broadcast from a heavily followed Twitter account would lead to a proportionally large spike in sales. Amplification, after all, comes from size: The more followers a person has, the more people who will see a message and who will, potentially, retweet it — and, thus, the more people who will potentially act on it. We know it intuitively: In general, the greater the numbers, the greater the viral power.
None. Literally, not a one. In fact — insult, meet injury! — in the days and weeks following Milano’s tweet, the book’s sales actually declined. The actress’ follower numbers, in this case, hadn’t been a force for much of anything. “At least with respect to the influence of behavior,” Christakis noted, “these links — these Twitter links — are weak.”
But, hey, maybe it was just an Alyssa Milano thing: It’s pretty fair to figure that the overlap between her followers and the universe of people who might buy a sciency book by two professors would be, you know, low. So Christakis and Fowler asked Tim O’Reilly — nearly 1.5 million followers, with, ostensibly, more book-interest overlap — to send the Connected link out to his feed.
The result? “We sold one extra copy of the book.”
If you’re interested in the way information spreads online — and if you’re interested in the future of news, you probably are — then the low volume-to-impact rate the authors found (which, though completely anecdotal, flies in the face of so much conventional wisdom) is fascinating. And it begs a question that appears so often in academic inquiry: What’s up?
In a talk yesterday evening at IBM’s T.J. Watson Research Center in Cambridge (we wrote about another IBM event, with dataviz guru Jer Thorp, this summer), Christakis, a professor at both Harvard Medical School and its Faculty of Arts and Sciences, dove into that question, discussing the particular (and peculiar) ways that social networks — online and off — work.
The talk focused on the epidemiology of action — how and whether certain behaviors spread through a population. (More on that here.) Though we often talk about social connections in terms of simple binaries — friend vs. not-friend, weak ties versus strong — the ties that bind people together, Christaskis’ research suggests, are nowhere near as simple as we often assume. There’s the obvious — your Facebook friend may not be your friend friend — but also, more murkily but more fascinatingly, the complex of connections that affect our behavior in surprising ways.
For the Lab’s purposes, one especially intriguing element of the discussion focused on Twitter — and the extent to which ideas spread through Twitter’s network catch on and have impact. One binary that might actually be relevant in that regard, Christakis suggested: influencer versus influence-ee. “If we’re really going to advance this field, we need to figure out how to identify not just influential people, but also influenceable people,” the professor noted. “We need not just shepherds, but sheep.” And “if we’re going to exploit online ties,” Christakis said — say, by creating communities of interest around news content, and potentially monetizing those communities — then “measures of meaningful interactions will be needed”: We need metrics, in particular, to determine “which online interactions represent real relationships, where an influence might possibly be exerted.”
For that, he continued, “we need to distinguish between influential, or real, ties online, and uninfluential, or weak, ties online.”
The next question: How do you do that? How do you look beyond standard (and, per Christakis’ anecdotal evidence, misleading) metrics like Twitter follower/Facebook friend counts and find more meaningful metrics of influence? One benefit of social networks’ movement online is that their dynamics are (relatively) easily trackable: We’re able as never before to put data behind the interactions that define society as a whole, and, in that, understand them better. (Connected, on the other hand — whose conclusions are based on data sets of social flow that were cultivated, over a period of years, from physical documents — didn’t have that luxury.)
And while Christakis’ talk raised as many questions as it answered — we’re still in early days when it comes to measuring behavioral influences online — one of his core ideas is an insight that several news organizations are already putting to practice: the power of the niche. Much more significant and influential than single celebrities — individual nodes in a network — are the “niches within the network where you have the particular assemblage of influential people and their followers.” When influence is layered — when its fabric is made stronger by tight connections across a smaller network — it’s more predictable, and more powerful.
And that has big implications not only for news organizations, but also for the platforms that are hoping to translate their ubiquity into financial and social gain. If you want your work to have impact, then targeting a bundle of closely connected networks — with news, with links, with messages — may make more sense than going for numbers alone. Spreading a conversation is not the same as affecting it. “I’m not saying that Twitter is useless,” Christakis said, “but I think that the ability of Twitter to disseminate information is different than its ability to influence behavior.”