The Ethics of Klout – Should We Trust In the Almighty Algorithm?

In recent months companies like Klout and PeerIndex have captured the imagination of the media by promising to deliver meaningful measurements of peoples’ online influence. But there’s a problem with measuring influence the way Klout promises to, and it has to do with the algorithms that scour the web to determine an individuals influence rating (or “Klout Score”, or “PeerIndex”).

The problem? The algorithms are secret.

(Of course there’s another problem, which is that Klout and PeerIndex don’t seem to spend a lot of effort trying to define “influence”, nor do they seem to get it right in their scores, see this video of a “Klout identified sheep expert” being interviewed. I’m not interested in getting into the nuts and bolts of measuring influence just now, although I have spent some time trying to twist my brain around how it might be accomplished. My argument is more basic, and it would stand even if it turned out that we could agree that online influence were measurable.)

If you’ve never heard of Klout, or haven’t gone through the process of setting up an account, here’s the gist of it. Anyone can create an account on these websites, link it to their Twitter and Facebook accounts (and other social media sites, though no Google+ or Blogs yet), and receive their influence score. Presumably, Klout and PeerIndex scour your tweets, retweets, mentions, direct messages, Facebook likes, and so on, in an attempt to see how many other people (and which people) are interested in what you have to say online. But at the end of the day, you are given a score that is meant to indicate how influential a person you are online. The score is visible to anyone who decides to look, and is even served up alongside every Twitter users’ avatar if you decide to turn that feature on.

These sites also suggest areas of knowledge that people are influential in. So you might turn out to have a Klout score of 45, and be told that you are influential when it comes to “medicine”, “privacy” and “cookies”.

When discussing the scoring algorithms each company has an offering. According to PeerIndex:

The PeerIndex algorithm recognizes the importance of speed and quantity by which we spot, share (and thus approve) content on any specific topic. Our choice of content recommendations can thus be used as a proxy to measure our knowledge, or authority, in a specific subject area.

Klout uses slightly more abstract language in describing their scoring algorithm:

We believe that influence is the ability to drive people to action — “action” might be defined as a reply, a retweet, a comment, or a click. We perform significant testing to ensure that the average click-through rate on links shared is highly correlated with a person’s Klout Score. The 25+ variables used to generate scores for each of these categories are normalized across the whole data set and run through our analytics engine. After the first pass of analytics, we apply a specific weight to each data point. We then run the factors through our machine-learning analysis and calculate the final Klout Score. The final Klout Score is a representation of how successful a person is at engaging their audience and how big of an impact their messages have on people.

Klout and PeerIndex are banking (literally) on the belief that these scores will translate into sales. How? By convincing people that Klout scores measure online influence, and translating that into a large membership.

For now, those sales seem to be driven through advertising partnerships. Klout offers its members Klout Perks, special offers for people who are influential. According to their website, “Klout believes that influencers deserve to be treated special.” For example, recently Klout members were offered exclusive “invites” to the highly anticipated new music streaming web service Spotify.

For now it also seems that all Klout users are equal in the eyes of the advertisers, but you can imagine a future where web influence, meted out by the Klouts and PeerIndexes of the world, translate into tiered perks: 50% off coupons for the low scorers, free iPads for the big influencers.

That’s the one side of measuring web influence, you can treat it as a means of building membership and selling stuff. It’s the side of influence that might bother some, but that most of us are used to. Influential people just get more free stuff. Nothing new there.

What is new, and problematic, is the way that influence a la Klout, is measured by an algorithm that only insiders get to tinker with.

Today, influence is quite a tricky thing to pin down, but mostly, influence has to do (or ought to do) with the fact that some people have demonstrated, usually publicly, that they are actually influential. That can mean that they are successful writers, actors, business people, or thinkers, who have proven that they have knowledge that is meaningful, and useful, to groups who then confer a certain authority in the form of influence onto them. Those groups might be peers or colleagues, or an audience that has access to the person’s activities that make him or her influential. The onlookers get to decide who is influential and who is not, and each new person has to decide, based on word of mouth from an other, whether or not to agree with the claim to influence. In deciding on influence in this way, open and socially, we can trust that the influence is deserved or not.

With online influence measurements like Klout, the power to confer influence is artificially concentrated in a secret algorithm. There is no word of mouth, no public display, no real point at which an individual, say you or I, can decide whether or not to confer influence onto another. The score is simply delivered.

By measuring influence this way, Klout is asking you to trust the Klout Score. But what reason do they offer for you to trust them? None as far as I can see. They claim to have sophisticated mathematical formulae at work, sure. But this is not a reason, it’s a way of trying to convince that the people at Klout are smart, and know what they are doing. It’s a way of saying, “just trust us” without actually explaining why we should.

If Klout were just about selling stuff then, like I said, it might not be such a big deal. But the other side of measuring online influence has to do with the influence. There is a real possibility that people’s Klout Scores will be taken seriously and treated (to Klout’s or PeerIndex’s joy) as an actual indication of their online influence, or even an indication of their professional abilities in general.

If you think this prospect is crazy, think again. Forbes recently interviewed a recruiter who suggested Klout scores could be used in hiring decisions. Microsoft also published a study that drew similar conclusions, finding that 89% of HR professionals consider online reputation when deciding on job candidates. Numbers have a way of becoming useful, even if they’re not meaningful. What could seem useful about Klout is that it gives the appearance of filtering through all the noise in people’s online lives to deliver up a nicely packaged report on that individual in the form of a number and some knowledge areas. Klout scores could be used as tie-breakers in situations where all other things about two individuals seem equal. In short, Klout Scores and PeerIndexes have the potential of benefiting some people over others in ways that go beyond the simple, trivial, benefits (coupons and what nots) of membership in a club.

There are predictable consequences if Klout Scores and PeerIndexes become useful as means of benefiting individuals in non-trivial ways. One of those is the potential for abuse, and corruption. Being in control of an influence measure, one that is somewhat impervious to scrutiny given its proprietary nature, will tempt a corporation and the individuals in it to trade in influence. (I say “will” instead of “could” for what should be obvious reasons to anyone.) Another is the potential for hacking the algorithm. Anyone successful in finding ways of artificially increasing their scores will appear instantly influential, and could receive benefits accordingly.

We need not jettison the idea of measuring online influence. But we ought to proceed slowly, and cautiously, before adopting any measure as meaningful. If online influence turns out to be important, and it may well be given the increasingly central role that online communications play in our lives, then we will want to set standards for measuring it. Klout cannot be that standard. Standards for measuring online influence can’t be determined behind the closed doors of corporations. They need to be transparent and open to criticism and debate. Without transparency, we have no reason to trust online influence scores, and less reason to use them other than in the most trivial of ways.


2 thoughts on “The Ethics of Klout – Should We Trust In the Almighty Algorithm?

  1. I just watched an interesting TED video about our increasing reliance on algorithms.

    It seems that, as our world becomes increasingly complicated, we become more and more reliant on algorithms to help us sort out the mess–or, at least, get what we want out of the mess. The problem is that we don't understand the systems we are applying the algorithms to in the first place, and the algorithms themselves are compounding the problem because they start to compete and interact in unpredictable, at least in a pragmatic sense, ways. It's sort of like playing pharmacist roulette; we don't know which medications are contraindicated.

    How do we handle the unfathomably intricate systems that are emerging from our heaps of data while, at the same time, having to manage algorithms which may-or-may-not have disastrous effects?

    Looking forward to your class in the Winter!

    1. Ryan,
      Yes, it is definitely a pressing challenge to understand the various ways that algorithms impact values, and to develop ways to mitigate the negative impacts and enhance the positive. I'll definitely take a look at the video. Thanks!

Leave a comment

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s