Researchers at the University of Michigan say they have developed an algorithm that is better than humans at detecting fake news stories.
Veronica Perez-Rosas, Rada Mihalcea and Alexandra Lefevre of the University of Michigan engineering and computer science department and Bennett Kleinberg of the University of Amsterdam psychology department found that humans were able to spot hoax stories 70 percent of the time while their algorithm was able to spot fake news 76 percent of the time.
{mosads}“You can imagine any number of applications for this on the front or back end of a news or social media site,” Mihalcea said in a statement. “It could provide users with an estimate of the trustworthiness of individual stories or a whole news site.”
The researchers used linguistic analysis to look at grammatical structure, word choice, punctuation and complexity to parse out fake news.
They honed the algorithm by feeding it fake test articles created from real news that had been altered into a fake story — a common method used by online hoaxers.
The algorithm could be a good sign for technology companies like Facebook and Twitter who have been fighting against the rapid proliferation of false stories that have spead across their platforms.
They have focused their efforts on using machine learning along with human reviewers, but haven’t been able to completely stop the spread of false stories yet.
But while the study shows algorithms can be more accurate and more efficient than human reviewers, its 24 percent fail rate suggests humans may still need to play a strong role in catching hoax stories.
Executives from Facebook, Twitter and Google are set to testify on how their platforms have been manipulated by foreign adversaries in a Sept. 5th Senate Intelligence Hearing. Countries such as Russia, and now Iran, have used a range of tactics in their attempts to influence the American public, including the dissemination of fake news stories.