Page 9 - My FlipBook

fake information is still going viral and spreading rapidly.” Me-
60
lissa Zimdars, an assistant professor of communication and me-
dia at Merrimack College agrees that Facebook’s fact-checking
system to combat fake news is falling short. “My initial read on
it is it’s ultimately kind of a PR move. It’s cheap to do. It’s easy.
It doesn’t actually require them to do anything.” Others point
61
out that these companies make money through advertising and
by generating more user engagement with content—regardless

of whether that content is true or false. Critics say this revenue
model gives them little incentive to shut down fake news, espe-
cially if it decreases their profits.

Artificial Intelligence

The sheer volume of fake news spreading across the Internet
makes it impossible for human users to find, identify, and remove
every fake story, picture, and video. Therefore, several tech com-
panies are developing computer algorithms and artificial intelli-
gence (AI) tools that can be used to identify suspicious stories.
One such company, AdVerif.ai, began public testing of its new AI
software in 2017. According to the company, the software can
detect fake stories, nudity, malware, and other unwanted content.
It does this by scanning content and searching for common signs
that something is wrong, such as a headline that does not match
body text or a headline with too many capital letters. (This is simi-
lar to how people can recognize phishing e-mails by the number
of misspelled words and grammatical errors.) The software also
cross-checks each story with a frequently updated database of
thousands of fake and real stories. The company provides its cli-
ents with a report for each story the software has reviewed and
assigns a score that assesses the likelihood that it is factual.
In testing, the AdVerif.ai program correctly classified articles
from the Onion as satire. The software also identified a Twitter
account that used a logo but promoted links that were not as-
sociated with the brand it was portraying. It correctly identified a

58