Klout is a popular Web site for measuring social media influence. You give it your social media details and it gives you a score using some top secret metrics. It also scans your networks to find topics you might be "influential" about.
When it's done, you can encourage the real humans in your life to help you adjust that score and those topics by awarding you points in certain areas. It's great - just when we got it out of our heads that we should be focusing on the number of followers we have across networks, this site has given us a whole other number to concern ourselves with. But can any amount of coded metrics - however well devised - ever actually give an accurate picture of something such as influence?
I was browsing Klout yesterday, minding my own business and giving friends influence points in topics as varied as “mobile apps” and "unicorns" (as one does) when a notification popped up. This is what it said:
"Klout believes you are influential about Terrorism."
Sure enough, "terrorism" has been added to the list of topics the site's engine believes Tallulah Habib might be somewhat of an expert in, alongside new media, Broadway, Linux and (oddly) cigarettes (I don't smoke).
What the tweet?!?
Now, perhaps my tweets over the weekend about the mainstream press' sensationalist Lonmin headlines were a bit impassioned, but does that make me a terrorist? Or is it my name, the conspicuous "Habib", that the site thought an indicator of possible terrorist tendencies? The thing is, I'll never know. If it was a human, it would have to justify its claims somehow; as a string of computer code, it's pretty unaccountable.
This is an extreme example. I know (well hope) that by now ITWeb knows I'm not a terrorist and the impact that such a claim by a non-sentient computer program could have on my life is negligible. I doubt the USA will add me to a "no fly list" just for this.
You don't need to be a member of Klout for it to run its analysis of you.
Tallulah Habib, social media activist, ITWeb
However, I think it raises valid concerns. Already, employers are looking at social media profiles to suss out employees before they come for interviews. Like it or not, our digital identities are being constructed. For instance, you don't need to be a member of Klout for it to run its analysis of you. You just need to somehow be connected to someone who is. If that concerns you, you don't even want to know what Google has already decided about you.
Increasingly, our online identities are more and more out of our control.
A common argument brought forward whenever privacy concerns are mentioned is: "I have nothing to hide." But that only applies if the data collected on you is accurate.
Way too big
The problem is there is too much data for human beings to sift through and the amount is only increasing. Big data - that's what they call it. So big that even the tools we used to use to mine data for us are ineffective, so we have no choice but to turn to computers to do it for us. Already all basic social media tools include some form of "sentiment analysis" for the benefit of online reputation management.
I was amused the first time I tried one for ITWeb's accounts. We'd been publishing a lot of stories about government tenders recently and those stories had been re-tweeted. The analysis tool thought, as I'm sure can only be expected, that this meant our audience felt "tender" towards us.
In the year-and-a-half since I tried that tool, the technology has grown and developed. The importance of context has been acknowledged, and as far as I can work out, most tools now attempt to take it into consideration. How can a computer ever fully comprehend the nuances of being human, though? Could a computer ever really understand sarcasm and humour when even we get confused at times?
The systems like Klout are a novelty now, but we're heading towards a world where they're the norm - as normal as Google. The US government might not be tracking my activity now based on a Klout score, but the idea is not outlandish. Michigan State Police can already scan mobile devices, scrape all the data off them, and quickly sort through the contents to identify signs of possible crime (including terrorism).
Perhaps the robot wars we've feared for so many years will have nothing to do with the likes of Skynet and The Matrix, and more to do with classification errors based on incorrectly constructed digital identities? Now that's a terrifying thought.
Share