Researchers at Stevens Institute of Technology are creating an AI-based tool to combat the scourge of fake news around the corona virus.

K.P. Subbalakshmi, AI expert at the Stevens Institute for Artificial Intelligence and a professor of electrical and computer engineering, along with Stevens graduate students Mingxuan Chen and Xingqiao Chu, are developing an AI-powered, scalable tool to detect and flag “fake news” relating to COVID-19. (News Medical)

COVID-19 “infodemic” could be lethal

 “Misinformation costs lives,” warned the WHO and UN on online rumors and fake news related to the COVID-19 pandemic. “Without the appropriate trust and correct information … the virus will continue to thrive.”

To control false rumors, panic and deaths due to medical misinformation it was necessary to flag media and social posts that constituted “fake news.” But the sheer size of the task made a manual approach impractical.

An AI solution to combat fake news

Dr. Subbalakshmi and her colleagues initially collected 2,600 articles on the subject of COVID-19 vaccines published over 15 months by as many as 80 different publishers. Each article was then labelled as reliable or undependable after cross-referencing it against the reputable media websites.

The next step was to gather over 24,000 Twitter posts that referred to the indexed news reports to develop a “stance detection” algorithm that could understand whether a tweet was supportive or dismissive of the relevant article.

Once their labelled data sets were created, the researchers developed a new AI architecture that could pick up typical, “giveaway” linguistic clues found in fake news, as different from general reports.

“It’s possible to take any written sentence and turn it into a data point – a vector in N-dimensional space – that represents the author’s use of language,” says Dr. Subbalakshmi. “Our algorithm examines those data points to decide if an article is more or less likely to be fake news.”

The algorithm is also able to pick up other pointers such as bombastic or emotional language, the time of an article’s publication, its length, even the number of authors and judge on an item’s reliability.

This system could detect fake news with about 88% accuracy, a marked improvement over other AI-based tools for combating fake news.

“We’ve created a very accurate algorithm for detecting misinformation,” Dr. Subbalakshmi said. “But our real contribution in this work is the dataset itself. We’re hoping other researchers will take this forward, and use it to help them better understand fake news.”

Related Story: According To AI, “Samson And Delilah” In The National Gallery Is Fake