April 26, 2011 by jimstogdill
Digital Tramp Stamps – Your friends are tattoos you may regret
I read somewhere that the single most important predictor of an individual’s obesity risk is whether or not their friends are obese. Eventually we can count on Facebook’s implementation of machine vision, weight estimating algorithms, and the publicness policies to justify them. That will be super awesome because it will save us all the trouble of updating our weight on our profile manually. We won’t even have to own tweeting scales to know just how obese are friends are.
In such a perfectly data rich world, Facebook’s voracious crew of ever curiouser Wants will be able to use the edge-strength-weighted average mass of your friends to easily predict your future displacement too. Where your friends are you will stochastically follow.
Just imagine your sense of wonder when you are innocently reading a well-paid blogger on Huffington Post and bam! An ad for a globo gym pops up in the margin. Why will that instill “a sense of wonder” you ask? Because you are two screening and 30 seconds later an ad for the exact same globo gym appears on your television too. Whoa! And the crazy thing is just this morning you were standing in front of the mirror thinking “Dammit, I wish Tim Ferris wasn’t such a tool, because I’m almost as fat as my buddy Dave and I really need to do something about it, but I can’t read that 4 Hour Fitness horseshit.”
Throw in politics, interests, hobbies, religion and every other thing about you that can be inferred from your cluster of friends, along with some inexpensive computational time to do the predictions and there they have it, your temporally distributed propensity functions for anything and everything you might spend money on. This my friends is what Chief Marketing Officer’s sacrifice chickens for. The truly killer app. Prepare to be pwned as they narrowcast their stream of perfectly timed and personalized sway at each and every one of your exact moments of want. We can haz yur wallet.
It’s not that easy? Fine. But I can say this with certainty: while you’re checking your Klout score (again) and reflecting on just how influential you are – a beacon of irresistible intelligence, wit, and wisdom – what’s really happening is your friends are influencing you. After all, you’re just one tiny voice to them, but they are a veritable din to you and they’re making you fat, politically angry, and probably high. Plus it’s all being conveniently digitized for later wantnalysis.
We like to think of social media as a platform for our personal expression but its a bi-directional conduit of ideas, expectations, affiliations, and all of the other grist of human social interaction. Part experiential and part aspirational the links say who we are and who we want to be. It’s a self-curated mesh of meme-sharing and world view that transforms us from an “I” to something more meta, the less differentiated “we.” We are so interleaved it’s impossible to know the vectors of cause and effect and influence. But regardless, the cluster centered on each of us has both descriptive and predictive power.
The other day LinkedIn told me that I should refer a few of my connections to my current employer. Apparently companies can send LinkedIn copies of their open job requisitions now and they’ll use machine learning to match your contact’s bio’s with them. Two of the referrals LinkedIn suggested used to work here so there must be something to the underlying logic. Apparently a naive bayesian algorithm and a bunch of commodity compute is all it takes to suggest that you might be a fit for a job you used to have.
That’s pretty cool but eventually I think they’ll do more than just match bio to req. Why not rank order the referral list by the probability of a successful fit – and do it by looking at the referral’s cluster of contacts and the attributes that are influencing them to either be successful, loyal, and dedicated, or lazy, shiftless, and job jumping. If Facebook’s social graph can predict what you’ll buy, LinkedIn’s should be able to predict how you’ll do.
Now my Mom was just like your mom when she said “I don’t want you hanging around with Mark and his stoner friends. Can’t you meet some nice people?” Well, now the Internet knows what your Mom knows – if you hang around with losers you are likely to turn into one, and conversely if you take a more aspirational approach and hang out with, well, Amy Chua’s kids, you’ll probably be awesome.
Fortunately back when I ignored my Mom it was before the Internet and there are no records of my bad choices. It used to be cool like that. Unless we were stupid enough to get arrested, the evidence of our transgressions decayed with the drug-diminished half-lives of our more stupider friend’s memories.
You, I’m sorry to say, aren’t so lucky. The Internet never forgets (Hadoop is an elephant after all) and it’s not just the picture of you puffing with Michael Phelps at some random party, but your relationship with that freak of nature itself is immortal. Your social errors preserved forever in the edge of a graph. Every connection you make with your loser, stoner, and stupid friends accretes another bit of ink to the web that once just influenced you, but now tattoos you.
You are your social cluster, and like that semi-ironic sleeve of barbed-wire ink you wear, it isn’t going away.