It's a wide, wonderful world of Twitter...but how much of it is "real"? This might seem like an ironic question to pose towards an online platform for a social until you learn that the 'twitter bot industry' — devoted to the creation and propagation of fake accounts — is a multi-million dollar enterprise. Some bots are funny, but most are simply a pain in the keester.

Corporations have also flocked to Twitter for promoting their brands, with what could be described as mixed success. With all the fakes and marketers out there, it'd be nice to have an easy way to tell if a new follower is a genuine human being.

Enter Gabriela Tavares and Aldo Faisal, two computer scientists from Imperial College London who have cracked the code and figured out an algorithm that distinguishes between human, corporate, and cyborg twitter accounts. The details of their work were published today in PLoS One.

The key to their method: timing. The researchers studied the timestamps for over 160,000 tweets from known personal, corporate, and bot accounts. They tracked between 70 and 90 accounts for each user type.

Tavares and Faisal predicted that certain patterns existed in the timing of twitter posts for each type of user, which could then be used to tell three groups apart. And they were right.

For instance, real humans liked to tweet from 7 a.m. to midnight almost everyday of the week, while corporations mostly tweeted during the workday and rarely on the weekends. Bots were erractic and sent out a high, consistent volume of tweets all of the time.

Based on these trends, they were able to create a computer model that could accurately guess — 80 percent of the time — whether a user was a man, a machine, or a business.

Arguably the coolest thing that the scientists uncovered was an equation that could predict when one of these users would send their next tweet.

Because Twitter places restrictions on how much data can be collected from the site, the team developed an open-source software called Twitter Reality Miner that scooped information from the media platform, but within its prescribed rules. It is publically available so others can replicate their work or find new patterns in Twitter usership.

It's hard to tell how these findings will impact the Twittersphere. The work provides lessons on human decision-making and how we decide to portray ourselves over an electronic broadcast, but also provide clues on how to mimic human behavior. Programmers could adapt the algorithms to devise better 'bot-blocking' Twitter apps, but on the other hand, people might use the research to build bots that seem more like humans, given there are already projects geared toward this goal.

"The identification and classification of specific types of users on Twitter can be useful for a variety of purposes, from the computational social sciences, focusing advertisement and political campaigns, to filtering spam, identity theft and malicious accounts," concluded Faisal.

Source: Tavares G, Aldo Faisal A. Scaling-Laws of Human Broadcast Communication Enable Distinction between Human, Corporate and Robot Twitter Users. PLoS One. 2013.