The term fake news has become popular in recent years, mainly due to the impact that disinformation campaigns on the net have had on various political and social events, such as the Brexit referendum.
A study on the use of Artificial Intelligence to detect these false news on Twitter published in the international magazine IEEE Access, analyzes mathematically what characteristics present the tuits that contain false information, and proposes a computer system to detect fake news.
The focus of the investigation is different from the usual journalistic approaches of fact checking, which require an enormous amount of work to verify data and review the newspaper library. Although Artificial Intelligence can help in this task, analyzing the content of the tuits automatically is very complicated, as it requires studying whether the author is simply being ironic or is really trying to pass off a false news as true.
Capital letters and icons are key
For this reason, in this work the researchers decided to use, apart from the content of the tuit, the data available about it and its author (i.e. the metadata), such as the number of followers at the time of publication, the date of registration in the social network or the amount of capital letters and icons used.
Research reveals that fake news on Twitter tends to include more polarized, novel and shocking information in order to attract the reader’s attention.
Experiments have shown that users who intentionally distribute false information behave differently than normal and this anomalous behavior is manifested through easily measurable properties, such as the number of contacts or tweets marked as the author’s favorites.
Consequently, these features can be used to assist in the automatic identification of potentially false news.
The paper uses Twitter data on the 2016 U.S. presidential election compiled by the authors themselves. The research findings were also presented at the Truth and Trust Online (TTO) conference recently held in London and organized by, among others, Twitter, Facebook and the University of Cambridge.