Can Machine Learning AI Help Us Discern What is Fake and What is Real?

April 19, 2018

254

April 19, 2018 – Facebook bots and fake news contributed to the current investigations in the United States around the 2016 election cycle. Discerning eyes may have been able to tell the difference between fake Facebook accounts and real ones. The same can be said for planted news stories that were blatant falsehoods. But for the undiscerning, it’s clear that there was a lot of pulling the wool over the eyes going on then, and continuing to today.

Creating a machine learning AI tool could be the best way to combat this online phenomenon. It might even become a downloadable app that we can all use to weed out so much of the garbage being spouted by hackers and cyber miscreants.

In recent weeks I’ve come across two attempts to create applications that can help all of us tell the difference between a friend request from a fake Facebook or other social media account, and planted fake stories.

The social networking machine learning algorithm was developed by researchers from Ben-Gurion University in Israel and the University of Washington in the U.S. The algorithm when finding fake social media accounts on one app correlated them to additional social networking apps. For example, it could discern that if it found a fake Facebook friend it was likely to find the same fake identity on your Twitter, Yelp, Instagram, and others. Testing was done on 10 different networks. They explain how the machine learning AI works in a paper published in the journal Social Network Analysis and Mining this month.

As for others machine learning AI developments, Aaron Edell, Co-founder and CEO of Machine Box Inc., invented an AI algorithm with better than 95% capability of knowing the difference between a real news story and one that is fake. Combining natural language processing and exposing the algorithm to real and fake news. A great source for the latter was The Onion, the iconic, satiric, online source of comedic fake news. For the real news, Reuters proved a useful training tool. With the machine learning AI exposed to thousands of news stories, it was its ability to recognize the real stories that ended up being the key to discerning the remainder as fake.

The fake news application is now available. It is called Fakebox and you can download it.

So what does the app do? Fakebox checks the following aspects of an article.

Title – titles can be assessed for bias.
Content – articles are analyzed to determine if written like real news or not.
Domain name – the origin of the news story is often a key indicator of its veracity with known domains that propagate fake content a key to discerning the reliability of the content.