The online network of Twitter and bots
https://www.bbc.com/news/technology-45918826
Social media provides an excellent way to keep up with the latest news and to join in on conversations about current events. But the problem is that these conversations and news can be artificial. Twitter recently released data about twitter bots and troll farms, revealing many of the techniques they use to manipulate public opinion. Some such techniques include being aggressive, or arguing on both sides of controversial topics in order to gather attention. It is somewhat disturbing how easy it seems to warp people’s sense of reality online, but it is also fascinating that such a phenomena is possible. It really emphasizes how important it is to be careful and skeptical online about not just the specific content that you read, but also about which conversations are fake.
The methods used by these twitter bots showcase the page rank algorithm. In twitter, one can use hashtags to label one’s tweet with keywords or topics. When people search up a word or topic, twitter returns tweets labeled with these hashtags, or even just tweets that contain those words. There are many criteria that can determine which tweets get priority in the search results. One is the number of retweets, which is the number of people who “linked” to your tweet. And this is the criteria we’ll talk about today. Consider what would happen if twitter were to use the same page rank algorithm we learned about. If there were to be an organization or a party of bots on twitter with the intent of artificially raising the importance of their tweets, they could try retweeting each other. Say that there are 100 bot accounts, and define a node or account as pointing to another node if it retweets or follows that node. 99 of them could point to several celebrities and prominent twitter users in order to raise their hub scores. These 99 hubs could then point to the remaining one account in order to raise its authority score. The one account could also use the techniques mentioned in the article, such as being aggressive or joining controversial conversations, in order to gather attention and incite reactions from real, authentic users in order to further raise its authority score. Thus this one account’s tweets could become higher priority in relevance. Now, thus twitter actually use this exact algorithm? Probably not. But the fact that the bot twitter accounts existed and succeeded in manipulating public opinion showcases how such a thing is possible, even if the method used to exploit the algorithm might not be exactly the same.
The twitter bots also allow us to imagine how twitter might look as a directed graph, with retweets being an edge from the re-tweeter to the original tweeter. If these bots really did use the method I described in the upper paragraph, i.e. having 99 accounts point to famous people and the one other account, then it’d be easier to spot groups of fake accounts. If one fake account were to be discovered, then many of the links to that fake account would be even more fake accounts. Groups that just retweet within themselves to artificially raise their relevance would also likely show up on the directed graph as a strongly connected component, or as a group of nodes with an unusually high number of inward edges compared to outward edges; such components of the directed graph could be ignored when running the algorithm.
There is essentially an arms race today involving many of the strategies discussed here, and much much more; websites want to keep their website from being exploited, and bots want to violate the integrity of discussions. Thus when you are online, try your best to stay skeptical of the things you read!