Taming fake news with logic and AI
The 23-year-old founder of UK-based machine learning platform talks about his team's experience of combing through 33,000 pieces of disinformation during the LS polls
The recently-concluded Lok Sabha elections turned out to be a tough task for Logically, a UK-based machine-learning platform that works on busting fake news. During the polls, a team of 25 in Bangalore combed through over 33,000 pieces of disinformation to put out analysed and credible news to users.
The platform was founded in 2017 by Lyric Jain, 23, a Cambridge and MIT student of artificial intelligence (AI) and is designed for gathering the biggest news stories from about 90,000 domains for determining the credibility of the claims in each article. Logically worked on the Swedish elections last year.
Massive flow of info
However, for the Indian-origin techie working on the Indian elections was an entirely new challenge. "Our platform can find and fact-check fake information in real time, and yet, we needed such a large team. You can imagine the massive flow of information in India," said Jain in a chat with mid-day on Wednesday.
Jain said a majority of Indian social media users who have been recently introduced to the Internet tend to be naïve while consuming and sharing fake information. "The largest circulated piece of fake information [during the polls] was about [Congress president] Rahul Gandhi being a foreign citizen. One more piece that became viral was about an actor's foreign citizenship," he said.
Speaking about how Logically was born, Jain said when he was in university, the Brexit debate had taken every news platform by storm. "People in messaging groups had extreme views. Information coming in from both sides — pro and anti-Brexit — was totally inaccurate, misleading and threatened the personal relationships. From this point, I started working on 'Logically' to help reduce the amount of misinformation and hyper-partisan information people were exposed to. We are now a fully-commercial platform offering free and customised services."
No political party here
In India, Logically has started working with various clients on a long-term basis, and in the past two months, worked with some on a case-to-case basis. The team includes data scientists, journalists, and AI experts who have built tools for offering a new level of insight into important news stories.
Has Jain worked with political parties? "Why would political parties hire us? In fact, a significant portion of fake stuff comes from them. We aren't a platform that tweaks algorithms to spread fake information. Our job is to filter fake information, fact check it and circulate its correct form."
The WhatsApp problem
For the team, filtering information on WhatsApp is a huge challenge. With 300 million users in India, the messaging app had become a campaigning tool. "It happens only in India. In countries like UK, people still use Facebook messenger and SMS. In India, we have other apps like Telegram as well. Since WhatsApp messaging is encrypted end-to-end, even the service provider cannot see the texts. This is the biggest challenge for fact checkers," he said. Jain wondered why WhatsApp hadn't thought of setting up its office in India till 2018, "Their campaign for preventing fake news is limited to 'stop sharing rumours'."
Busting fake info
But, Jain said, his experiment for checking WhatsApp rumour-mongering should help a great deal. "We have been able to create a reverse chain of messaging that offers correct information to the end users. For example, if fake information gets out from one individual to 30 others, we would be able to offer verified content to one person in that group, which they can share with members who have read the fake content. It should make the problem solvable."
Images shared on the platform remain an issue. According to Jain, text verification is comparatively easier, but images can be verified only through tools not all users may have access to. "Memes get shared more than most popular news articles," he said. To show how disinformation is spread through photos, Jain took out his phone and showed a scan of a union minister's resignation letter, which had a forged government logo.
For now, Logically is filtering memes in two languages, "We are focusing on Hindi and English, verifying the content and passing it on to regional feeders, who in turn, reach out to regional media," said Jain.
When it comes to reporting and commentary, Jain said news television in India has more commentary whereas the print media sticks to facts. He sees the print media as factually correct when compared to other formats, but added the major players had extreme biases.
Need correct info
As for tools that can make the lives of fact-checkers easy in India, Jain said they needed easy access to correct information, especially from the central and state governments. In UK and USA, he said, all kinds of information is readily available in a machine-readable format, "But in India, reams of PDFs are uploaded on the websites. If I were to ask about expenditure on a particular scheme in a particular district, I won't get the answer in real time. I will have to download pages, find relevant columns and do calculations on paper." Jain said this online information mess was a blot on the expertise of a country regarded as one of the information technology captains of the world.
Sign up for all the latest news, top galleries and trending videos from Mid-day.comSubscribe