Scientists can tell the difference between human tweeters, automated Twitter accounts and those managed by groups of people based on tweet timings.

The researchers, from Imperial College London, found that they could easily tell if humans were responsible for tweets, irrespective of their content, based on when the user’s tweet was posted. Their technique also enabled them to distinguish between individuals and those accounts managed by groups of people.

Surprisingly, they found that there was not an obvious pattern for automated Twitter accounts, which were more random than humans in terms of the timing of tweets.

The program could provide another way to verify tweets

The technology behind today’s research could provide a new way to verify the authenticity of Twitter account users. The study is also providing fresh insights into how our brains operate when using social media tools.

The study was published today in the journal PLoS One.

Dr Aldo Faisal, author of the study from the Departments of Bioengineering and Computing at Imperial College London, said: “In our study, we found that it does not matter whether you are a model in Sao Paulo or a student in Beijing; it seems that we all share something very similar in our tweeting pattern that dictates the timing of when we send out messages. Why is this important to know? Well, it suggests that there is something fundamentally similar in the wiring of our brains that comes into play when we use twitter to communicate.” 

Since its creation in 2006, Twitter has accrued more than 500 million users worldwide, producing approximately 65 million tweets per day. The popularity of Twitter makes it an important tool for journalism, marketing, political campaigns and for advancing social change.

The team analysed 160,000 tweets from around the world using computer programs that they developed to spot complex behavioural patterns in Twitter users.

By analysing the timing interval between tweets the team were able to determine that personal accounts holders tweeted more evenly throughout the week and, on each day, more tweets were recorded during typical waking hours of 7am to midnight. Managed accounts, run by people in corporations, were more active during the five working days and during working hours between 8am and 8pm. Tweets from automated accounts on the other hand were less easy to predict using their method because they were random in terms of when they were sent out.

The researchers believe that randomness in the timing of tweets from automated accounts is because they have programmed activities that vary considerably and that they tend to be reactive, responding to external events. This make automated accounts less uniform in their behaviour.

There have been several studies carried out before to analyse the contents of tweets in order to verify who it is we are communicating with. Despite a high success rate, these methods can be expensive to carry out.

The researchers say their computer program could provide another method for verifying twitter account holders, which could avoid complications in relation to processing different languages and understanding the context in which tweets are sent.

Dr Faisal concludes: “I am a Twitter user and this project was born out of my curiosity in wanting to know who I’m actually tweeting with. Sure, there are methods for verifying account holders, but are they failsafe methods? Are we really speaking to a Member of Parliament on twitter or a PR representative or a machine? We have now developed a method that can help us determine who it is we are actually tweeting with.”

The team, which also includes Gabriela Tavares from the Department of Computing at Imperial, have now made their computer program available for public use, which can be downloaded at www.FaisalLab.com.