Google-Backed A.I. Aims to Help Journalists Write Better News Stories
JUICE project sifts sources for inspiration.
Image: Sean Davis/Flickr
The web has distorted journalism, for better and for worse. Global events are covered instantaneously by multiple news agencies. Citizen journalists rush to promote news on social media. But the drive for constant content to make ends meet for publishers has also borne the rise of "churnalism", and demolished the traditional media ecosystem.
Clickbait, in particular, has had an undoubtedly negative effect on readers—the very people that journalists promise to set out to inform—by dishing up endlessly repackaged press releases that retain their PR distortions and untruths. And you only have to look to the US presidential elections to see the impact the constant obsession for reaction pieces and statistical exclusives has had on disorientated readers.
But to try and curb this downward spiral of content, one group of researchers has turned to artificial intelligence—with the help of Google.
Neil Maiden, professor of digital creativity at UK's Cass Business School, is part of a team in London that is building this tool. The idea is to help journalists create more unique, higher-quality articles by using the power of artificial intelligence and machine learning to suggest additional information for journalists to add to their work. But while AI is already writing news stories, can it help journalists build better stories?
The project that Maiden is heading up is called JUICE, and is funded by Google's Digital News Initiative to help journalists "discover and explore new creative angles on stories they write". A collaboration between Cass Business School and the Department of Journalism at City, University of London, JUICE's core product is an add-on to Google Docs that uses AI systems such as language processing, web searches, and recommendation algorithms to help journalists write their stories.
"Our intention is to support journalists very discreetly in the work that they do," Maiden told Motherboard. "As you're writing a story you can, with a click of a button, ask [JUICE] to invoke different ways of offering you creative and productive advice about the ways in which you can take your story in valuable directions."
"Our intention is to support journalists very discreetly in the work that they do"
The main engine of the JUICE tool is access to 470 news sites. Whenever a journalist types something actionable into the document, such as the name of a newsworthy figure or a subject, the AI automatically performs "creative searches" on these websites and pulls up relevant articles, cartoons, and other multimedia that can be included in the story. "If the user asked for X we will generate a query which is X + Y + Z, where Y + Z are things the journalist didn't know they might be interested in," said Maiden.
"The essence of the tool is six strategies," he added. "We worked with experienced journalists and asked them how they would elaborate or develop angles or spin a story in a particular way to make it interesting, unique, and relevant. Each of the strategies is implemented automatically. The app can take what's been typed already and process it and generate different kind of searches."
The six strategies identified were: individuals, causal, quirky, quantifiable elements, ramifications, and data visualisations. For example, with 'ramifications', the AI may help the journalist creatively explore information about the consequences of events and actions related to a news story in order to generate new and useful stories about future consequences. The 'quirky' search strategy sifts through a digital database of 30,000 political cartoons so the journalist can take the underlying themes or comedy of them to seek inspiration, Maiden explained.
The tool exists to help journalists, but a byproduct would hopefully be higher quality journalism. Maiden said that the tool had proven successful in training student journalists how to create more unique stories when it was trialed at City University.
"Before we took the earlier versions of the tool into professional working newsrooms, we thought we'd test it on the students. One of the first things that the staffed realised is actually this can first of all train journalists more quickly. The tool is enacting how experienced journalists work," said Maiden. "Hopefully, if the tool works effectively, it's pulling out interesting sources so you're getting more value, novelty, and surprise."
The tool is still in development, and Motherboard hopes to get access to JUICE early next year.
Get six of our favorite Motherboard stories every day by signing up for our newsletter.