Michael Wu, Ph.D. is Lithium's Principal Scientist of Analytics, digging into the complex dynamics of social interaction and online communities.
He's a regular blogger on the Lithosphere and previously wrote in the Analytic Science blog. You can follow him on Twitter at mich8elwu.
Last week, Fast Company launched the Influence Project, which claims to identify the most influential person online. I’ve written an article on this project last week: The Fast Influencer Myth. Since this topic garnered much discussion, I wanted to take an opportunity to address some of the more interesting topics that have arisen from the subject. I was going to start my mini-series on the business implications of what we’ve learned from the social anthropological roles of communities and social networks (loosely referred to as cyber anthropology) but I will do that later instead.
What’s Wrong with the Influence Project (IMHO)?
I think I was pretty clear on this in my last post, but I wanted to start this article by saying that in my opinion the Influence Project is not measuring influence at all. Influence is not just about affecting the behavior of those you interacted with, as Fast Company defined it. It matters a great deal how you affect their behavior. If you spam someone and frustrate him into clicking your unique link, that is not influence. Moreover, since Fast Company encourages participants to use any means to spread their link, there is no defense against users who may abuse and game the system. Technically, you can lie and trick someone into clicking your link. That is definitely not influence!
What I dislike about this project is that Fast Company is putting an attractive label, “the most influential person,” on these potentially misbehaving spammers, tricksters and liars. By featuring these people on their magazine, Fast Company essentially glorifies this misbehavior. This just doesn’t pass my “gut check” as being something good and worthy of promoting.
How Can I Fix the Influence Project?
Several people have asked me, how Fast Company can improve the methodology of this poorly designed experiment? Being a social analytics scientist who has investigated the topic of social media influence, I thought perhaps I could offer some constructive thoughts on the subject.
There are probably many things that I would have done differently if I was devising this experiment to make it scientifically rigorous, but that would probably take away a lot of the entertainment value too. So, to make the problem more well-defined, I gave myself a more challenging problem: If there was ONE thing I can change about the experiment, what would I do to make the experiment better?
At first I thought to myself, I couldn’t possibly turn this project into a scientific rigorous experiment under this constraint. However, after the World Cup final, when I was a little lubricated from cheap wine, I was daring enough to let my mind wander for a few hours over this problem. I’m pleased to say, that the wine did the trick, and I did come up with couple of changes that Fast Company could do to improve their Influence Project! And here they are.
Social Shame: A Defense against Deception
Although I believe that most people are good at heart and well intentioned, we all know that there will always be a small group of trouble makers who will misbehave, especially under fierce competition. One effective defense against deception is social shame. So how can Fast Company make one simple change to their experiment to put this into effect? Fast Company simply needs to store the tweets/updates from which a participant’s unique link has been clicked, and state that they will be displayed along with the photo of the winner in their November issue. That is it!
This simple alteration – the addition of a public audit trail – can dramatically change the psychology of the game, because not only will people see the winning contestant (the most influential person), they will also see exactly the means of how he won. Knowing this, participants who are truly influential and care about their reputation would likely refrain from using any deceptions, because any misconduct that earns points will be revealed to the public with complete transparency. For every desirable reward there need to be an associated cost. Otherwise the system cannot sustain itself economically. Likewise, when there is potential for social fame, complete transparency provides the mechanism of social shame to keep the misbehaved in check.
Normalized Accounting: A Defense against Spamming
Although social shame is a powerful deterrent against social misconduct, such as deception and manipulation, it does not however, guard against spammers. A simple way to defend against a spammer is to change the influence accounting. Basically, we need to normalize Fast Company’s influence score by the amount of tweets/updates.
Mark Borden (Fast Company editor writing about this project) said in his post “Someone with 100,000 followers who only gets 100 people to join the project is less influential than someone with 150 followers who gets 100 people to join.” My question is which of the following is more influential?
Someone who has shared his unique link 100,000 times over the course of this experiment and get 100 people to join.
Someone who has only shared his link 150 times and got 100 people to join.
If I use Fast Company’s overly simplistic definition of influence, then by the same logic the latter should be more influential. Why did they only address followers, but not frequency of sharing? Do they encourage spamming? Why? I’ll leave that for you to ponder.
To implement this simple change, Fast Company would have to keep a count of how many times each participants shared their unique link. Since each link is unique to a participant, this shouldn’t be difficult. Then divide their influence score by the shared-count to come up with a spam-corrected version of influence score. This completely changes the dynamics of the game. Now, spamming others with no effect can actually reduce your influence. So if people want to increase their influence, they better share less and share more effectively.
Although Fast Company’s Influence Project (or experiment) is fraught with errors, I was able to cook up a couple of small changes that could help to address the problems with deception and spamming. These changes will not correct Fast Company’s simplistic definition of influence and their flawed experimental design. Consequently, the experiment is still far from perfect, and it does not really measure influence in any rigorous sense, but it will be better than its current state. So what is the moral of today’s post?
Clever use of social shame can serve as a defense against deception
Measuring influence with normalized social accounting is less prone to spamming
A little bit of alcohol can help opening your mind. Don’t abuse it though!
Alright, I will see if there is anything that I can write about the Influence Project next week. Maybe I will have another spark of inspiration. Otherwise, I will resume our journey on cyber anthropology. In the mean time, let me know what you think. Was this the wine talking or was I enlightened? Kudos, comments, questions, discussions, and criticisms are all welcome as always. See you next time.