Right now, telling if a tweet is from an Internet bot or from a real, live person is pretty much a hit and miss thing at best.
Conventional wisdom says the same tweet sent to everybody is likely from a bot. A tweet replying to a tweet of yours in about a second is also a bot as is a Twitter account that follows hundreds of tweeters but only has a few followers. But these methods are prone to a lot of error.
Computer scientists at Indiana University have developed a tool that enables Twitter users to tell if a human or a bot actually operates a specific account. The tool is called BotOrNot. It joins a growing number of online tools that try to determine if it's a bot or a human behind a tweet.
BotOrNot looks at over a thousand features from Twitter's social network, including a user's Twitter posts and other information. The tool analyzes the account in real time and assigns a probability on whether or not a specific Twitter account is being run by a bot.
To develop the tool, Indiana U researchers looked at the habits of Twitter bots created by a Texas A&M University professor's infolab. They train statistical models so that they can tell the difference between humans and social bots.
Researchers claim BotOrNot is very accurate. BotOrNot gets a .95 score, with 1.0 indicating perfect accuracy, in a widely used measure for accuracy.
They pointed out that many bots are totally benign or almost harmless. There are, however, some nasty bots used to mislead, exploit and manipulate discourse with spam, misinformation, rumors, and malware.
Researchers believe these bots can be dangerous and have the potential to create panic and make it easier for cybercriminals to thrive, among others.