A polarizing election, economic uncertainty and international tensions may have contributed to undermining society’s faith in the accuracy of news. Further complicating matters is how that news is consumed—often from alternative non-verified sources, such as social media and search engines—where one person’s opinion or bias suddenly becomes “fact.” This shift in the media landscape has been exploited by some for political goals or financial gain, giving rise to the current “fake news” epidemic.
Efforts to combat misinformation in news have largely been ineffective. The reason? The massive amount of fake news stories in circulation (with new ones coming every day) makes it too time and labor intensive for a human to review each story. Despite with a system that allowed users to flag suspicious stories, the system—subject to bia and user error as with any human-driven system—was ultimately scrapped because user feedback from bad actors ended up taking down legitimate stories.
Technology cuts both ways; the same artificial intelligence (AI) algorithms that can predict a user’s personal preferences and deliver customized content are now being used to drown social media channels and websites in a tsunami of bogus news stories. And leveraging AI to identify fake news is not a simple proposition.
Helpful tips on how to spot fake news stories can be effective, but it’s not scalable. While some of the tips could be handled exponentially faster by AI in comparison to a human (for example, spotting a suspicious URL or grammar mistake), other tips (being able to recognize a fact as a lie) require levels of analysis and interpretation that are currently too difficult to automate via AI and machine learning. As one article illustrates, while AI is great at performing straight-forward tasks, the subtleties involved in spotting incorrect or misrepresented facts in a fake new story may be beyond the capabilities of today’s AI.
To unleash the full potential of AI and big data analytics against the fake news epidemic requires access to highly accurate data. In fact, data veracity is one of the four key pillars of big data (volume, variety, velocity, and veracity) because no matter how advanced a big data platform is, its results will be tainted if the underlying data is flawed. While other types of bad data may be incorrect because of errors (which can be found and corrected), false news stories are intentionally written to be incorrect and camouflage their falsehoods by mimicking the style and tone of legitimate stories, making them highly challenging to spot without human intervention.
What Do You Think?