We need to talk about identifying trolls....
Updated 15/11/2017 regar
Seems like magic right? This process of deci
The truth is, it isn't and the time has come that we talked about the complexity of identifying the trolls in our social networks.
"This combination of factors prompted me to run a case study experiment and see what came back. Also because people have been asking for an easy way to identify trolls."
Firstly, a troll is not a bot.
A bot is an automated account which is programmed to respond to keywords or to retweet. They have no human involvement aside from the programmer. They can also be rea
A troll, however, is an account manned by a human and subsequently they will more than likely pass undetected. Many of the accounts I've been investigating over the last few days are troll accounts and not bots. Identifying just one is a laborious and complex business.
Thankfully, a case study popped up in my timeline this morning, respon
So there I was, having a coffee and min
What struck me, however, was "Wot". In almost a year of daily interaction with Scottish Twitter users, I can hand on heart say I'd never seen one of them use "Wot".
This combination of factors prompted me to run a case study experiment and see what came back. Also because people have been asking for an easy way to identify trolls.
So, was this just an unpleasant in
As with everything troll related, it's complicated.
First of all we have no information on this person to go on, barring a Twitter alias. However, even starting with nothing always leads somewhere.
Through a handy piece of analytical software which integrates with Twitter's API data, that one piece of information soon linked @
In fact, there are 388 twitter accounts
(accounting for followers and follows) linked to
So, yes, it's the same person.
What's also interesting is the organic growth of the account over the two time periods. While a number of common connections remain, there have been substantial changes in follow patterns over both iterations.
Taking a closer look at the current persona, it was also quite clear that they tweet at some volume compared to their actual following - which is a pre-in
So, with the warning signs all there, a sample of the timeline starts to shape a picture of troll behaviour.
Smoo, as the account calls itself, doesn't really talk about the things many others are
In the first iteration of
In fact, they used to be an even more overt retweet account for Russian memes and what we now know to be Russian and Russian-inspired
So, there's a pretty convincing case
But, like most troll accounts, they are seeded with local information, which makes them very hard to
Their use of the word "Way" to replace "With" is not Scottish. In fact, having confirmed this by consulting a broad spectrum of Twitter users from across Scotland, "Way" would almost certainly not be used. Rather, "With" would be substituted with "Wae" or even "Wi".
N.B. One of the ways troll farms study native languages is by watching local me
However, there was ad
Establishing whether this is legitimate proof of identity is again not straight forward, but rewar
Identifying a troll account is, as you can see, incre
In this case study, the account features a number of classic in
On balance, at the higher end of the probability scale, it was a fair assessment that this account appears to be a foreign-based troll pushing Russian messaging.
But you have to make up your own mind, that's the thrust of all this. And the danger.
When challenged, their reply was as follows:
An alleged owner of the account has since approached a Scottish me
The complexity of identifying troll accounts is a long overdue
As further food for thought, I'll leave you with the tweet from Sputnik last night which resulted in my timeline being filled with a number of accounts like
In the mean time, and in short: when it comes to trolls you are going to have to make your own judgment calls until Twitter gets its act together.