As data analytics company, at A4E we are more than familiar with the capabilities and potential of Artificial Intelligence widely known as AI, especially combined with some kind of automation. As society members, just like you, we also know that the fake news is not just pieces of information heavily disliked by politicians with disputable reputations. This is why we are curious whether AI is capable to help users and media businesses to distinguish the real news out of fake ones.
Why is such task important? Indisputably, the fake news was turned into media celebrities by Donald Trump and his call for alternative facts. We have to admit that the fake news was existing way before his election as US president. Fake news is capable not just to shape the public opinion in particular direction but also to be a fruitful profit source for the click-bite websites. The biggest harm they bring is blurring the visible line between good and bad.
So, let’s go back to the initial question – is AI capable of stopping fake news? Such hypothesis should be examined by defining what fake news is and to train AI to spot it. An experienced journalist can identify the fake news in a moment or after a few minutes research. Unfortunately, the regular information consumer lacks the journalist` background and knowledge and a research of his own will take more time.
AI can be helpful if is filled with particular and well-defined tasks. Some researchers as Rohit Akiwatkar gives 4 different factors helping AI to distinguish fake out of the real news:
- Web page scoring
- Facts identification
- Reputation prediction
- Sensational words identification
He is presuming that if a web page is relatively new with a low page rank scoring and the facts stated in the piece can not be find in other news stories and the reputation of the news outlet is relatively low and there are some sensational words it is more likely to classify this particular information as fake news. There is a logic behind such presumption and an AI can perform such operations.
What if the web page is relatively new web media, the fact is a unique finding of the journalist and his media, the reputation is still low because the media is still new and there are some click-bait words in order to generate more web traffic easily? This does not make the news story fake!
Such case is a type 1 and type 2 errors widely known in data analytics. Everything suggests that something should be classified as one thing but actually, it should not be. This is where AI cannot beat the human brain. If it was able, AI would be as good as actual human intellect.
Few more pieces in this puzzle. How to make AI distinguish the fake news out of the parody websites with made up news just for fun? There is one more – it is natural to make a mistake no matter you are a professional journalist or not. AI cannot distinguish the honest mistake.
Recognizing what’s fake requires not just the kind of pattern recognition that AI is so good at, says Cade Metz in an article for Wired. It requires human judgment, as Dean Pomerleau, adjunct professor at Carnegie Mellon acknowledges. A machine that can reliably identify fake news is a machine that has completely solved AI. “It would mean AI has reached human-level intelligence,” says Pomerlau.
It is known that AI simply cannot provide the ultimate answer whether something is true or not, whether an article is misleading or not. This is why AI alone can’t beat the fake news.