Twitter has hit one of its lowest points after the proliferation of dirty campaigns, fake news, and other content directly tied to politics, especially in the United States. Some profiles go around spewing information, and the first assumption is that they are programmed bots. Twitter isn’t doing anything to separate real users from artificial ones, so two UC Berkeley students decided to respond with a Google Chrome extension called Botcheck.me, which evaluates profiles in an attempt to determine whether they belong to a bot.
The Battlefield of Post-Truth
For some time now, Twitter has abandoned its role as a social network to become a political battlefield. There isn’t a public official, leader, activist, or militant who doesn’t have an account there, plus journalism dedicated to capturing everything they post. The quality of the message or its truthfulness no longer matters. The idea is to repeat it tirelessly, get it everywhere, and be passionate in the process: the mechanics of “post-truth” in its purest form. Twitter also enabled a change in the rules, allowing these profiles to post 24 hours a day without pause. It’s not hard to imagine an account having more than one administrator, but there’s another resource available: bots.
Unfortunately, Twitter doesn’t perform any kind of active identification that lets us separate bots from real accounts (it’s not exactly simple to do so), and that’s where Botcheck.me steps in. It’s a Chrome-compatible extension developed by UC Berkeley students Ash Bhat and Rohan Phadte. Using an advanced machine learning system, Botcheck.me seeks to expose bots that distribute political propaganda. This means its AI model isn’t very good at detecting bots that share other types of content, but if something is related to Donald Trump, Hillary Clinton, Republicans, Democrats, or Russia, it will most likely find them.
Twitter insists that bots account for 5 percent of its platform, but external researchers like Bhat and Phadte believe their presence is ten times greater. When it comes to accuracy, the number they share for Botcheck.me is 93.5 percent. Even so, the challenge is far more complicated than it seems. Some bots can respond to private messages, and there are always hybrid profiles that combine bots with human intervention. It might not be very useful for us, but Botcheck.me is simply necessary these days.
