AI to the rescue!

Fake news now claims a disturbingly large chunk of the news we consume. And in all likelihood, you wouldn’t have noticed it. But its impact on brands, personalities, markets, issues and causes has become significant. Media studies experts say fake news has influenced two recent, epochal events: Brexit and the US Presidential elections. If you want an example, a study from New York University and Stanford has found that people shared fake stories supporting Trump at least 30 million times on social media during the presidential election campaign. In comparison, fake pro-Clinton stories were shared at least 7.6 million times. Back home in India, which seems to have become the new hub of fake news, misinformation spreads like wildfire, damaging companies and individuals alike.

And that’s why digital marketing firms, corporates and news organisations are now brainstorming frantically to filter and fume fake news. Granted, it’s not an easy task given the challenges fake news poses to human intelligence, in terms of identifying and classifying sources of news, fact-checking claims and verifying quotes and numbers. But it seems humans have found a suitable ally in artificial intelligence, which can help us trace and trash fake news.

Armed with Big Data analytics, AI works in many layers here. At the outset, it does what it does best: find patterns. Most fake news follow similar patterns. They sensationalise even trivial information, skip citing sources for numbers and are first beamed from websites that lack credibility. AI tools can sift through millions of webpages in real time and set off alarm bells if they detect what could be cooked up news.

That said, how ‘exactly’ does AI work against fake news? For starters, there are a few tangible steps. It starts with rating webpages. The AI programme will run a check on the news sources’ URL and try to analyse their reputation. Of course, an original news item from, say, a Financial Times is far more acceptable than a news report from an unknown portal that is produced in one of the content cottage industries in Macedonia. As the algorithm gets perfected, news organisations can build a repository of trustworthy news sources and the rating process will get fine-tuned and fast.

The next step is crunching numbers, which is especially important in business news. Umpteen number of news items appear on the web detailing false financial performances of companies. This news pumps half-baked data. AI helps analyse these numbers and puts them in perspective and finds correlations that help us ascertain their reliability.

Language mapping is the next crucial stage in filtering fake news. AI tools help detect unwanted sensationalism and wordplay in news and alert readers. Analysts have observed that fake news makers are mostly amateur content creators working for money or pushing a cause. They rarely manifest restraint when it comes to language use. AI tools, especially those with NLP (natural language processing) capabilities, can help here.

AI also helps in areas such as stance detection. That means to scan the copy and find whether the author or the agency that has reported the news is in favour of or against the target of the news. This inference will help trace ulterior motives, say experts.

Facebook, which is one of the most popular social media platforms where fake news spreads, is already using AI to fight the menace. The other two big players, Google and Twitter, are also developing and integrating AI plus Big Data tools into their information dissemination infrastructure. Other than the internal programmes of the big-ticket social media companies, several other small players have developed AI solutions that help check fake news. NewsWhip, a social media monitoring firm from Dublin, is helping several media companies sieve out fake news.

Crowdtangle, a content discovery firm, also offers similar services. GoogleTrends, Hoaxy, Pheme, Snopes are some of the entities that offer AI-Big Data solutions to fight fake news.

Even the academic world has sat up and taken note of fake news woes.

In the US, the WVU Reed College of Media has tied up with the computer science department at the WVU Benjamin M Statler College of Engineering and Mineral Resources to host an AI Vs fake news course at its Media Innovation Center. Several other universities are following suit.

Experts expect AI in checking fake news will see more R&D activity given the way governments are pushing for clean news. Germany, for one, has approved plans to penalise social media companies more than $50 million if they post fake news.

But one of the biggest challenges these companies face is time. Analysing fake news in real time is a big challenge. It requires highly potent machine learning skills and rapid-fire analytics for data veracity. Here, the companies hope crowdsourcing will help them significantly. Communities and developers spread across the globe are now developing tools, small and big, to track fake news and alert readers and news media. Big players like Google want to help compile and coordinate these efforts.

Another worry is of the algorithm going wrong. AI experts say there are possibilities, statistically speaking, for AI tools to produce two kinds of bad results — false negative and false positive. Simply put, this means an AI tool could stamp a fake news item as not fake and term real news as fake. But that’s just an initial hiccup. As we move on, and with more and more data and news getting cleansed, we will soon be able to usher in a world free of fake news. And, hey, that’s not fake news!