Can algorithms really tackle the fake news fiasco?

A study published by the University of Oxford pegs the 2016 US presidential elections as a “watershed moment” that witnessed social media manipulation at its worst. Trump and Clinton camps, for instance, harnessed big data and bots to swing public opinion in their favour.

Closer home, Mumbai witnessed widespread paranoia triggered by a social media hoax that claimed the city was directly in the path of cyclone Phyan.

Ex-defence minister, Manohar Parrikar found himself at the receiving end of a social media scuttlebutt that claimed he would go back to the Centre if he lost the Goa assembly bypoll.

And the fake news hotchpotch is not confined to spurious news websites or Facebook alone. In India, a major chunk of hoaxes circulate on WhatsApp. Although WhatsApp is equipped with end-to-end encryption, a lot of messages are recirculated due to the sender and receiver having unique sets of keys. This makes it hard for authorities to trace the message back to the original perpetrator.

Given the humongous reach and impact of social media, companies are now harnessing AI-based algorithms to keep fake news at bay.

Leading the charge in the battle against fake news is Google. The internet behemoth has taken the menace by its horns. Its PageRank algorithm ranks Google search results based on the pages that are linked to the publication. What makes the cut is that PageRank applies across multiple platforms, and therefore gives us the ability to identify fake news from real.

Another way to zero in on fake news is Google’s reverse image search. Users can upload the image or photograph in question, and Google pinpoints the source of the image, and websites on which it appeared.

Facebook, like Google, is going all guns blazing in combating the fake news threat. Facebook did not rely solely on algorithms at the very beginning, though. The company hired a team of news curators to compile and build its ‘trending news’ section.

In an interview with Gizmodo, many of these news curators revealed that they were actually hired to “train” Facebook’s algorithm. The curators also alleged that Facebook instructed them to not mention their job roles in resumes or on public profiles.

In 2016, Facebook replaced its team of news curators and editors with its AI-based algorithm. And the results weren’t very encouraging – Facebook’s algorithm failed to identify the hoax about Fox News firing its moderator Megyn Kelly after an alleged “treachery”.

However, Alex Stamos, Chief Security Officer at Facebook justified the need for using algorithms. He recently tweeted that in any situation where millions or billions of items need to be sorted, algorithms have to be used.

Nobody of substance at the big companies thinks of algorithms as neutral. Nobody is not aware of the risks.

Human intervention is of course more reliable than algorithms. This is because algorithms can be manipulated or influenced by trolls and lack judgement when it comes to complex editorial decisions.

Although algorithms are less expensive, highly scalable, and easier to maintain, they tend to fail unless continuously trained and retrained on humongous volumes of data. They are heavily reliant on volume and momentum of data. The technology is still in its nascent phase, though it hides huge potential for publishers.