Twitter Could Help Predict Riots, Study Suggests

Researchers in the Middle East say Twitter could be a useful tool in predicting violent riots.

The Qatar Computing Research Institute says that although you can't read too much into a single 'tweet'  from a single user, when you gather the data from thousands of users you can detect certain trends.

By piecing that data together, researchers say they can predict when and where violence will occur.

To test the theory, the Qatar researchers developed what they call the 'Political Polarization Index'. They analyzed all Twitter posts by Egyptian users and gave each user a rating between 0 and 1 based on whether they were more or less likely to share content involving religious or secular public figures. (Source: qcri.qa)

Hashtag Diversity Linked To Unrest

The next step was to look at hashtags, which are phrases Twitter users add to posts about a particular topic that make it easier to search and track discussions.

For each hashtag the researchers calculated whether it was equally likely to be used by "religious" and "secular" posters.

According to the researchers, the higher the disparity between the two sides, the more likely it was that political unrest would break out. In fact, they found a Twitter discussion foreshadowed violence last November and December relating to Egypt's new constitution.

One Man's Coup Is Another's Revolution

In some cases, the clues came from the two sides talking about a similar topic, but using different terminology. For example, secularists were more likely to talk about the most recent change of government as a "revolution," while religious posters described it as a "coup".

One limitation to the analysis is that, by the time the disparity is noticeable, tension is already so palpable that unrest is virtually impossible to prevent. (Source: newscientist.com)

It's not the first time online activity has been used to predict major events. Google tracks searches relating to flu and flu-related symptoms and says it can use that data to predict potential epidemics. The firm insists it's a  quicker and more efficient method than more traditional surveys by health professionals.