Some Fake News Fighters Embrace AI, Others Seek the Human Touch
Fake news doesn’t seem to be going away anytime soon, and some entrepreneurs are targeting false news reports with new services designed to alert readers.
Some countries have pushed for new laws to criminalize the creation of fake news – raising questions about government censorship – but these new fake news fighters take a different approach, some using Artificial Intelligence, some using human power, and some using a combination of AI and humans.
Several high-profile fake news fighting services have launched in recent years, some of them driven by the amount of fake news generated during the 2016 U.S. election. These services generally focus on web content appearing to be legitimate news, as an alternative to traditional fact-checking services like Snopes – which takes a broad look at Web-based news and rumors – or PolitiFact – which addresses claims made by politicians and political groups.
The amount of fake news generated during the election campaign was the main reason FightHoax founder Valentinos Tzekas began working on his service two years ago. At the time, Tzekas was a first-year applied informatics student at a Greek university, but he is planning to leave school to work full time on FightHoax.
The 2016 U.S. elections “took the world by storm,” said Tzekas, named to the Internet Society’s 25 under 25 list in 2017. “All of a sudden, rumors and fake news started coming out of nowhere.”
Tzekas saw news reports about a student Macedonia making thousands of dollars each month by writing false news stories. “The worst thing is that people believe anything they read on the Internet,” he said.
FightHoax takes a tech-centric approach to identifying fake news by using Artificial Intelligence, including IBM’s Watson, to rate articles on seven criteria. The algorithms test for the quality and level of the writing, whether the article includes polarized language, and whether the headline is clickbait. The service also checks the political leaning of the publication, among other things.
While thinking about fake news, “one day I thought to myself: ‘Can I make something to analyze news articles and warn readers about anything suspicious, such as the use of propaganda rhetoric?’ Tzekas said. “’Can I make something that will, in a few minutes, do the work of human fact checkers and get a result as to whether a given story is true or a hoax?’”
FightHoax, which claimed an 89 percent accuracy rates in early tests, is planning for an enterprise dashboard version release within 18 months, Tzekas said. He plans to work first with newsrooms and academics, with the enterprise dashboard allowing the service’s API to connect with advertising-serving companies, news distributors, and social networks.
The advantage of an AI-powered approach is a service that can “analyze harmful news content that exists on the Internet, at a scale,” Tzekas said. While a tech-focused approach doesn’t work to analyze all factors involved in fake news, it’s a good fit to identify markers like clickbait headlines, he added.
“Machines cannot fully understand the messy-polymorphic human written language so sometimes you need to think really simple,” he said. “Our number one mission here at FightHoax is to make people think. [We want to] solve disinformation at a scale, with technology that works on a human scale.”
At the opposite end of the technology spectrum from FightHoax is NewsGuard, announced in March. The service, cofounded by journalism veterans Steven Brill and Gordon Crovitz, will take a human approach to rooting out fake news, with plans to hire dozens of journalists to review and rate 7,500 news and information websites most accessed and shared in the United States.
The trained journalists will write “nutrition-label” style reviews of the sites and include green, yellow, or red labels. The founders plan to license the service to social media platforms and online search companies, as well as to interested consumers.
In April, NewsGuard launched a fake news hotline for the public to report suspected sites.
If FightHoax embraces technology to fighting fake news, and NewsGuard embraces a human approach, U.K.-based Factmata splits the difference. The service, with several high-profile investors, uses a combination of AI tools and human intervention to target hate speech, propaganda, and spoof news sites.
The combination of AI and human insight promises to be the best way to identify fake news, said Dhruv Ghulati, Factmata’s founder and CEO. “Pure human is too slow and unscalable for the pace of today’s news cycle, and pure AI just will get a lot of things wrong if not carefully supervised and trained,” he said.
Factmata has, in fact, reached out to NewsGuard about a potential partnership, he added. “Their community of journalists we feel can be greatly augmented by our pipelines for assessing content and judging news for its credibility, and our AI to help automatically flag certain types of content,” Ghulati said
The target audience for Factmata’s news platform is journalists and other experts, but the company is also planning a B2B product that provides news quality scores for customers including advertising networks and agencies.
Factmata plans to soft launch its Web-based Briefr news sharing service within weeks, with about 350 journalists and other experts participating, Ghulati said. A full launch is scheduled for June.
Read our interview with Wired Editor in Chief Nicholas Thompson on the changing role of media, then explore the 2017 Internet Society Global Internet Report: Paths to Our Digital Future and read the recommendations to ensure that humanity remains at the core of tomorrow’s Internet.
The post Some Fake News Fighters Embrace AI, Others Seek the Human Touch appeared first on Internet Society.