One of my earliest blog posts was “A Twitter Analog to PageRank”, in which I proposed a simple measure for Twitter influence.
As the title indicates, the measure is inspired by PageRank. It relies on the insight that Twitter users — at least the human ones — have a finite supply of attention, which they distribute among the accounts they follow. The more accounts a user follows, the less attention for each account.
For simplicity, the model assumes that Twitter users divide their attention uniformly among the accounts they follow, and hence that the probability of their seeing a particular tweet is inversely proportional to the number of accounts they follow. It models retweeting by assuming that users retweet any tweet they see with a constant probability p.
The result: a recursive formula that measures a user’s influence by modeling the expected total number of people who will see a tweet from that user:
My readers decided to dub the measure “TunkRank”, and the name stuck. I challenged those same readers to implement the measure, and Jason Adams stepped up, implementing the measure and hosting it on a website.
But that was only the beginning. Social networking researchers noticed the blog post and began making their own implementations and citing it. A search on Google Scholar today reveals over 200 references to TunkRank, many of which are in ACM or IEEE publications. I even scored a press mention for it.
In retrospect, the model for TunkRank is shockingly naive. It relies entirely on the follower graph, which means that it doesn’t take into account the quantity or quality of tweets. Moreover, it doesn’t incorporate any engagement signals. Nonetheless, this simple measure demonstrated promise in research studies.
I’m gratified that TunkRank has made a small contribution to the research literature. But I’m even happier that I was able to make this contribution through a blog post and the ensuing collaboration with an informal social media community. Indeed, TunkRank was a breakthrough for me personally, after which I truly appreciated the power of sharing ideas through blogging.
I’m also happy that I was able to get so much mileage out of such a simple idea. It seems completely obvious that any measure of influence would need to take into account that people have a finite supply of attention. Yet, by stating the obvious, I managed to make a small but constructive contribution to the community of people studying social network influence.
I haven’t thought much about influence measures in the past few years. But I continue to share my ideas, and I hope that people still find them useful.