“The Three A’s” of Identifying a Twitter Bot
By: Account Executive Brook O’Meara-Sayen
In my last blog piece, I discussed what a Twitter bot is, provided an extremely basic overview of how they’re made, and discussed how Twitter bots can make and change sentiment online. Much of the recent social discourse regarding bots has been negative, mainly due to revelations that Russia utilized a veritable army of bots in 2016 in their attempt to influence the US Presidential Election. Russia used its bots to move online sentiment, creating the impression that hashtags campaigns and other orchestrated social media content were coming from actual voters, and not a shadowy office building in St. Petersburg, Russia. To do this, they relied on the assumption that the everyday American Twitter user wouldn’t be able to tell the difference between a bot and person. In many cases, they were right. Russia also had a dedicated team of ‘professional trolls’ working in tandem with the bots, making it hard to discern who was mechanical, and who wasn’t. However, most bots are still relatively easy to find–and ubiquitous. An estimated 9-15 percent of all Twitter accounts are bots.
So, how is one able to identify a bot in your feed? Attempting to weed out bots isn’t foolproof, and the tips I’m about to give will not always work. They will allow you to analyze and think critically about the suspicious accounts you may come across. One of the easiest ways to identify a bot is by looking at post frequency and identity, by using “The Three A’s”, a system coined by the Digital Forensics Research Lab.
THE THREE A’S
- ACTIVITY, or how much do you post?
Machines are great because they can perform menial tasks much faster than any human ever could, just ask Henry Ford. This is in part why bots who retweet original content are so pervasive on the web. You can create a ‘retweet bot’ in a matter of minutes. Once it’s on, it won’t turn off unless you tell it to. This leads to a twitter account with an abnormally high number of tweets – the first red flag. The Digital Forensic Research Lab treats any account that tweets more than 40 times a day as suspicious, and anything over 140 as highly suspect.
- ANONYMITY, or who are you?
Creating a convincing fake online person can be tedious, so most bots tend towards vague anonymity. They might use generic names, false locations, and minimal or misleading bios that lack personal information. Most human twitter accounts will include at least cursory identifying information, such as a verifiable name and profile picture. They may also tweet identifying characteristics out about themselves unknowingly, such as a picture of their dog, child, car, etc., or a complaint about the weather or commute in a specific location. This is not to say that all anonymous twitter accounts are bots, but used in conjunction with other warning signs, anonymity can be a helpful indicator.
- AMPLIFICATION, or what are you saying?
Bots cannot easily create lucid, fully-formed thoughts on a subject. They rarely provide the nuance needed to trick a human user. So, what do they do instead? They cheat. Bots might post content written by real people. They retweet, copy news headlines verbatim, and you hope no one questions why there is no obviously original content on their page.
Using the three A’s when looking at a suspected bot account can give a user a good sense of its authenticity, but it is not foolproof. In addition to the three A’s, users should look for other indicators such as stolen pictures, accounts with very few followers and insanely high engagement rates, and usernames that appear to be randomly generated.
It’s hard to tell what’s real and what’s fake on the internet. If you’re interested in becoming a full-fledged bot-finder, I would also point you to some of the source material for this blog post, “12 Ways to Spot a Bot” by the Digital Forensics Research Lab.