One of the most loaded terms in today’s political climate is “fake news.” Not a day goes by when, for example, the president of the United States doesn’t attack some unfavorable-to-him news story as a specimen of “fake news.”
You might dismiss such comments as a ham-handed attempt to divert attention away from President Donald Trump’s massive credibility gap. Indeed, the underlying fake news issue is vastly overblown, as it almost certainly was during the 2016 US presidential campaign. But you can’t deny that people’s wariness regarding fake news is not going away anytime soon. Just as you can’t deny that political operatives—both domestic and foreign—are doing everything in their power to manipulate the daily news cycle to their advantage.
The “fake news” problem is overblown
AI has been implicated in the fake news problem. There is little hard evidence that political operatives have used the tools of the AI trade to build or distribute their bogus content. But that hasn’t stopped many people from adding fake news to the Pandora’s box of evils that natural language processing, convolutional neural networks, and other data-driven algorithms have supposedly unleased on the human race.
Looked at in the broader perspective, it’s clear that AI could be a contributing factor to fake news. But it could just as easily be a part of the solution. After all, every side of the political equation now has access to these tools, so maybe AI is simply a double-edged sword that we’ll have to get used to using.
To what extent is AI a culprit in the fake news issue? No one doubts the technology’s ability to fabricate every possible type of content, using generative techniques to produce the most astonishingly authentic-seeming text, images, video, audio, and even live conversations. That algorithmic prowess will continue to improve, thanks to the growing abundance of training data, sophistication of generative adversarial networks, and availability of skilled data scientists the world over.
As this consciousness seeps into the popular mind, those of paranoiac bent will continue to believe there’s a vast conspiracy of fake news purveyors, even when—as with say, “alien abduction” stories—there’s next to no hard evidence for it.
How AI can help address the “fake news” issue
But, of course, AI can be a powerful tool for flagging and possibly blocking fake news before it pollutes the popular mind. Already, we’re seeing the best data-science minds address this problem head-on through open-source community challenges, research projects, and commercial solutions. Obviously, social media powerhouses and content publishers are investing heavily in this technology.
Much work remains to be done, but you can already see the outlines of how AI can provide strong defenses against fake news.
Fact-checking is of course the heart of it. That’s easier said than done, of course, due to the obvious fact that there’s no unimpeachable source of truth to which algorithms may defer on every conceivable topic, especially for breaking stories in which fresh facts may be coming to light for the very first time.
In those scenarios, AI’s ability to do anomaly detection can be a key fact-checking resource. If a specific news site remains the only one reporting a supposedly hot story after several cranks of the news cycle, there’s a good chance that this is fake news (or it simply might be satire, sarcasm, stupidity, or something else this side of blatant deception).
Likewise, AI-infused natural language understanding can deconstruct the semantic components of a specific story, and then cross-check the accuracy of each against reputable sources. Furthermore, predictive AI models might be used to quarantine suspect news stories for real-time vetting by human fact checkers. Crowd-sourcing might even be used to provide a scalable pool of on-demand fact-checking resources to feed training data to AI fake-news-detection algorithms.
Discrepancy-checking is another fake-news-sniffing process that AI can do much faster, more scalably, and accurately than people. To the extent that fake news purveyors are skimping on editorial processes such as proofreading, AI algorithms can rapidly cross-check for grammar, spelling, syntax, punctuation, and other errors in fake news content. In addition, AI computer vision algorithms may be able to spot fabricated photos and other doctored images in bogus stories.
Spam-checking is another AI-powered capability that can prove useful in holding back the tide of fake news. To the extent that botnets—and even legions of political partisans—are sharing, liking, and otherwise propagating bogus stories on the internet, one might train AI models to detect this in real time so that these activities don’t swamp the news cycle before it’s too late. People may be manipulated by sensationalistic planted stories to do just this, so an AI-identified fake-news-throttling capability might be absolutely essential to prevent such social engineering attacks from doing their damage to the news cycle. Already, the anti-fake-news community uses some defenses—such as bogus-site blacklists and junk-content flagging add-ins—that got their start in the antispam arena.
Of course, none of these safeguards will stop partisans from eagerly subscribing to every fake news site in the world that supports their cherished illusions. Likewise, there’s little we can do technologically—without strangling free speech—to weed out authentic news stories that include deliberately fake “facts” spread by political operatives.
All of this points to something we should all have learned in civics class growing up: Democratic societies depend on citizens who are well-informed about public affairs. This, in turn, requires that each of us open our minds to many sources of news and that we think critically about them. Just relying on your Facebook feed to keep you plugged into the world at large is a surefire path to ignorance.