Roses are red, violets are blue, fake-news-detecting AI is fake news, too

Humanity's bulls*** is too much for software

What could you do with fifty grand, though?

Another group tackling the same problem is Full Fact, an independent fact-checking organization in the UK.

Armed with a €50,000 grant from Google’s Digital News Initiative, Full Fact was one of three teams to win funding for tackling fake news.

“Fact checking is difficult. Everyone thinks it’s a matter of yes or no, but it’s not that simple. It’s complex, it requires a lot of nuance – something that computers aren’t good at,” said Mevan Babakar, digital products manager at Full Fact.

The human fact checkers over at Full Fact don’t label information simply as true or false either. Evidence supporting and undermining the claim are laid out and it’s up to the public to make up their own minds.

There are levels of complexity to the problem, Babakar explains. “For example, something like population numbers can be checked against data – that’s easy for computers and could be automated. But for claims like ‘the NHS is in crisis’ – that requires interpreting different datasets and meanings, so it’s not something a computer can do.”

“Human fact checkers come with a body of experience; they know the methodology behind the data and they know its limits,” she added.

Full Fact is turning away from the glitz and glamour of AI and machine learning and are instead focusing on customizing Solr, a search engine, with APIs that will collate information on repeated claims made over the internet or television.

The team uses natural language processing with search patterns and queries that monitor the spread of information and locate the primary source needed to judge the accuracy of information.

The search engine will power two Full Fact tools: Trends and Live. Trends works similarly to Google Trends: it reveals who is repeating inaccurate claims, allowing Full Fact to quickly identify the spread of misinformation and asking journalists responsible for the errors to make corrections. Meanwhile, Live allows the charity to flag claims and factcheck them during parliamentary debates, or on TV, in real-time.

“People are much more attuned to misinformation now, because they’ve seen so much of it in such a short amount of time,” Babakar said. But unlike the Fake News Challenge, Full Fact is looking for short-term remedies that can be rolled out in the next six months.

“Machine learning and AI over promises but under delivers. It’s something I believe doesn’t require a five-year PhD in machine learning or neural nets. Some of the technology is pretty basic and we are already seeing it work in Trends and Live.”

“It’s an interesting space, though,” Babakar said. “I wonder how machines will deal with things like satire. One person’s satire can be another person’s fake news.” ®

Sponsored: How to get more from MicroStrategy by optimising your data stack

SUBSCRIBE TO OUR WEEKLY TECH NEWSLETTER




Biting the hand that feeds IT © 1998–2019