Facebook Friends Don’t Influence Each Other – Or Do They?

So it’s a great headline, ‘Facebook friends don’t influence each other‘ – or as Wired soundbite it, we’re immune to viral marketing. Except that it is wrong.

You may be reading reports with similar headlines, covering a new Harvard study “Social selection and peer influence in an online social network” appearing in this month’s “Proceedings of the National Academy of Sciences of the United States of America” by Kevin Lewis, Marco Gonzalez and Jason Kaufman that largely confirms some research we covered last year from Sinan Aral.

If you haven’t read it, then read on as your client or boss is quite likely to collar you with a challenge in the near future; ‘Didn’t that recent Harvard study show all that social sharing is nonsense?’ (read – stop messing around with social technology and do something more worthwhile with our money.)

Well no.  What this study did do is look at the movies, music and books people set as their favourites in their personal Facebook profiles, and then look to see over time whether friends’ favourites tend to come together (cluster).

Guess what? They don’t.  Set your favourite movie as Bladerunner in your Facebook profile, and your friend sets it as The Wizard of Oz – and its unlikely that any peer influence – coercion (use of force), persuasion (use of argument) or suggestion (use of example) – will result in switching favourites.

That’s it – that’s the study.  But a title ‘Stochastic modelling shows that people who set Bladerunner as their favourite movie fail to convince their friends on Facebook to do the same’ has a less of a ring to it.  So the study didn’t show we’re immune from social influence or viral marketing at all.

In fact, Facebook experiments suggest quite the contrary; ‘broadcast sharing’ (aka liking) generate a 246% increase in peer influence and ‘personal sharing’ aka ‘sending’ generate a 98% increase in peer influence, whilst another study in the Harvard Business Review (see below) on the effect of personal recommendations, showed that a $31,500 marketing campaign designed to activate recommendations boosted profits by $481,090

However, the latest Lewis et al study is useful because it did find that your choice of who you friend in Facebook in the first place does seem to be influenced by what your friends like – if your favourite movie is Bladerunner, then you’re more likely to friend someone else who has a similar favourite.  You’re also more likely to friend people based on similarities in location (physical proximity), gender, racial background, social class, and birthplace.  It’s the ‘liking heuristic’ in action – we tend to like people similar to (like) ourselves.

In other words, the study came down on one side of a great debate raging right now as to whether social networks – whether digital or traditional – amplify peer influence – or not.

What we know is that when you take a snapshot of people together – virtually or physically – they tend to be more similar than you would expect.

But is this similarity because we influence each other through social ties (in techno-jargon because of ‘social contagion’), or is it because our similarities brought us together in the first place (in techno-jargon ‘homophily’)?  There’s an increasing body of evidence that suggests that homophily (‘birds of a feather flock together’), not contagion (influence) drives much clustering.  What Facebook does is bring similar people together rather than provide a platform for people to influence each other.  The implications for social commerce and social media marketing are obvious; without the contagion effect, the value of social technology in commerce and marketing is considerably lessened.

But this study doesn’t show that contagion doesn’t happen – and whilst my PhD in social contagion means I have a vested interest in this debate, let’s not throw out the contagion baby – and a century of experimental results – out with the stochastic modelling bathwater.  And let’s be wary of misleading headlines.