According to Kallos, predicting the behavior of a crowd is possible through the detection of a decisive «indicator» amidst the «noise».
The idea of using social media, such as Twitter, as a means of forecasting future developments is almost as old as the social networks themselves. In recent years many have occasionally claimed that it is possible to predict election results, product trajectories in the market, successes (or failures) of movies, etc., through analysis of social media data.
One of them is Nathan Kallus, of MIT, who has developed a method for predicting crowd behavior, through statements and comments on Twitter. Specifically, according to MIT Technology Review, Kallus has analyzed the tweets related to the 2013 coup in Egypt, and claims that the whole unrest associated with the incident was clearly predictable, several days beforehand.
On a purely practical level, it is not difficult to imagine how the future behavior of a crowd can be reflected on Twitter, for example through reports and updates about gatherings and their coordination via it. Such activity on social networks is a clear indication of specific mass behavior in the future.
According to Kalous the prediction of a crowd's behavior is possible through the detection of a defining «indicator» amidst the «noise». As he notes, this is possible through the «fishing» tweets, with the aim of detecting references to future events and then analyzing the trends (trends) that are linked to them.
«The gathering of crowds and the execution of a common action can often be missed through trends that appear in the data beforehand» says in context.
With exactly this kind of analysis a company under the characteristic name «Recorded Future» is engaged, which scans 300,000 different «sources» on the Web in seven different languages with the purpose of identifying references to future events. Analyzing this data, Kalos seeks to predict phenomena such as large‑scale protests and riots. «We discover that the volume of publicly available information has the ability to reveal the future actions of the masses» he says.
At the first level, it is defined as «significant» a protest/unrest that receives more coverage than the conventional media usually does. Subsequently, this coverage is analyzed so that it becomes clear when such phenomena occur and the activity on Twitter that precedes it is sought. The goal is the identification of «patterns» so that corresponding «sequences» become perceptible in other cases. Regarding the fall of Morsi in Egypt, Callus believes that the riots, which started the chain of events resulting in his removal from power after a military intervention, were predictable via Twitter many weeks before their onset.
Source: naftemporiki.gr
