Delivering email at scale is a technically challenging endeavour, and getting it wrong could shut down the entire email communication channel. In fact, this…
It’s common knowledge that Twitter is home to more than a few bot accounts, be it for spam, promotions or political purposes. Now, a new study (via CBS) has found that up to 15% of Twitter accounts are actually bots.
If one of the more recent Twitter statistics of 319 million active users is taken into account, then that means that over 47-million users are bots.
The study, by researchers from Indiana University and the University of Southern California, saw a framework for bot detection being created. The result?
“Our models yield high accuracy and agreement with each other and can detect bots of different nature. Our estimates suggest that between 9% and 15% of active Twitter accounts are bots.”
In fact, the study cautioned that the 15% figure might be conservative too, owing to the possibility of sophisticated bots that escape human judgement.
How many bots are on Twitter? Well, a new study points to a possible figure of over 47-million
The authors noted that there were bots that served a useful purpose, such as disseminating news and coordinating volunteer activities. But it added that there were social bots that did the opposite, such as manufacturing fake political support, promoting terrorism and disseminating conspiracy theories.
The researchers said that their bot detection framework takes over 1000 features and six key factors into account. These factors are user-based features (number of followers, profile description, number of tweets), friends features (retweeting, mentioning, being retweeted/mentioned), network features (retweet/mention networks and hashtag co-occurrence networks), temporal features (average tweet rate over set period, distribution of intervals between events), content and language features (length and entropy of tweets) and sentiment features (emotions).
“We classified nearly 14 million accounts using our system and inferred the optimal threshold scores that separate human and bot accounts for several models with different mixes of simple and sophisticated bots,” the authors wrote in their conclusion.
“Training data have an important effect on classifier sensitivity. Our estimates for the bot population range between 9% and 15%. This points to the importance of tracking increasingly sophisticated bots, since deception and detection technologies are in a never-ending arms race.”
Featured image: Chris Corneschi via Flickr (CC By-SA 2.0, resized)