You didn’t believe me when I said, “In ten years nobody will be able to lie,” did you?
All I can say now is, “Ha!”
In the few weeks since I wrote that series, I have seen many articles touting progress in the area of truth-detection. Here’s the latest one, from The Atlantic: Algorithms Can Help Stomp Out Fake News. You can visit the link for the full story, but here I’ll give you a peek at just a few of the fascinating techniques that are already in use today.
Let’s say a fake news story appears headlined, Scientists Predict Mile-Wide Asteroid Will Strike Earth Late Next Year, and the Twitterverse goes wild. How might computerized fact-checkers determine whether the rumor is true?
To get started, a computer can do what you and I would do: consider the source. If the story appeared in a publication that is known at least attempt responsible journalism, such as The New York Times, then it is more likely to be true. If it appeared only in fringe publications, then it is probably false.
The article does not state this, but I’m sure that an automated fact-checker could understand that the story about the asteroid pertains to something that astronomers would know about, so it could check authoritative sites for astronomy.
What makes fake news so dangerous is that it spreads like wildfire across social media. People share it with each other and if it has enough of a “hook” it can go viral, taking on a life (and truth) of its own. Fortunately, it turns out that falsehoods tend to be shared differently than the truth, and computers can pick up on the differences.
In an algorithm reminiscent of Google’s PageRank, a computer can assess the likely truthfulness of a tweet by seeing how many followers the author has and how many tweets he has made in the past. The higher those numbers, the more likely it is that the person is not injecting fake news from the fringes of the Internet. So, is news of the impending asteroid disaster merely being spread from one person’s crazy uncle to another’s, or are non-fringe figures concerned as well?
The computer can refine its assessment by seeing whether the tweet is often retweeted with a skeptical comment. Computers are able to understand the meaning behind statements like, “But if it’s not going to strike for several months, it must be far away. We can’t predict the path of such distant objects with any accuracy.” Also, if a retweet contains a link to a fact-checking site such as Snopes, where the news is debunked, that’s a big red flag. If many people are skeptical, chances are we should be, too.
For those who care, it will be increasingly easy to sort the Internet wheat from the chaff. This is good news!