Does the Influence Engine weed out spammers?

Does the Influence Engine weed out spammers?

Our last blog post about the Top 40 List of #NEfollowers provoked an interesting comment from (check out the website, it”s pretty cool). Chris raised some really good points that we thought could best be answered in a new blog post.

Chris” comment begins like this:

Interesting, but I can’t help but think it’s also rather flawed.

For example, let’s say some leading light in community says “Lets all post our website addresses!” and the rest of the community accordingly tweets their URLs, your algorithm isn’t going to pick that up if it’s only looking at personally references, named-checks and RTs. That person is clearly quite influential though.

Here”s Dollywagon”s response to this interesting point:

We think people who have the ability to galvanise everyone else to do stuff like you suggest can usually do so for 1 of 2 reasons.

The first reason is that someone has had a blinder of an idea with a life of its own that makes perfect viral material. We see this stuff on the web all the time because the web is a very big place with lots of people randomly producing really mad stuff. The same can also be said about people who are first on the scene with an epic piece of news that everyone else wants to know about.

But it”s very rare indeed for a single individual to create a piece of viral content during their own lifetime that goes on to become a huge viral hit on the web. The Influence Engine couldn”t predict which “outlier” is going to release the next viral craze any more than the world”s best minds could predict the winner of next year”s Derby.

Nonetheless, you have given us an idea – we can actually track the appearance of URLs in #NEfollower hashtags and analyse the way they get propagated around the network. We”ll give this a go and see what happens.

The second reason why someone might be able to get lots of other people to follow their lead as you suggest would be because an individual already has a lot of credibility, prestige or authority in their network or community. These retrodbonus are exactly the type of people that the Influence Engine will pin-point because the pattern of their network relationships (or topological profile) will have betrayed the esteem in which they”re generally held.

Chris” comment continues:

Similarly, if someone is adding the tag too *all* their tweets and the community choruses for them to stop with “@name please stop spamming”, your algorithm is going to push the spammer to the up the influence ranking when really their actions should be pushing them down.

Without analysing the *context* of tweets, whether they represent a positive or negative weighting, I can’t see how this chart is particularly useful.

It is pretty though, and that’s something.

And here”s Dollywagon”s response to this perceptive point.

If a person does get “flamed” by a community in the way you suggest, they will of course get picked up very quickly by the Influence Engine. In normal projects we usually take regular snap-shots of a community and track the relative rise and fall of its members over time.

The most influential and respected members of the community tend to maintain their prestige over the long term, but the people who get “flamed” tend to shoot up and then crash down to earth like fireworks on bonfire night. This makes them very easy to spot.

As we continue to monitor the #NEfollowers community we”ll get to know the network gurus and recognise the spammers and flamers for what they really are.

Thanks also, Chris, for your compliment about the image – I”m pleased you like it. You might like to click through to this network image too.

2 People read this post and left a comment

  • Chris 10/06/2010

    I see. That’s really interesting. I’m looking forward to seeing how it pans out.

    A small aside: are your network map images made with Processing ( They have that look to them.

  • admin 10/06/2010

    No, we don’t use software – we’ve developed our own imaging applications which we generally output into Inkscape svg files for touch-ins.

Leave a Reply