For the news industry, the promise — or perhaps threat — of automation is that technology will be able to handle more of the monotonous reporting, freeing up human reporters to do the enterprising, high-value work.
Reuters, however, sees another path: cybernetic reporters. At NICAR on Friday, Padraic Cassidy, Reuters’ editor of news production systems, took the wraps off Lynx Insights, a new in-house automation tool designed to augment reporting by surfacing trends, facts, and anomalies in data, which reporters can then use to accelerate the production of their existing stories or spot new ones.
While Reuters has experimented with automated reporting since at least 2015, Cassidy said that the process was not only expensive and time-consuming, but often resulted in articles that were transparently written by a machine. “After looking at those stories, we decided to be sensible about it and made it so that machines can do what they’re good at and marry that to what humans are good at, which is judgment, context, quotes, and insight,” he said.
Via Lynx Insights, which Cassidy said was “much more sane, less expensive, and quicker-to-market,” Reuters reporters can easily surface key data related to their stories. A markets reporter, could, for example, use the tool to quickly determine the total value of merger and acquisition deals for a given year, or analyze historical trends in commodities pricing. So far, Reuters has focused on markets data, but Cassidy said that over time Reuters will expand to other kinds of “low-hanging-fruit stories,” particularly in sports and corporate earnings numbers.
Reuters reporters have had a direct role in building Lynx Insights. Reuters’s Toronto-based data science team made a callout to reporters to determine what kind of data they’d be most interested in and what kinds of stories took them the most time to produce. Speed, of course, is a prime consideration with automation, which is why newswire companies have embraced the technology more readily than other news organizations.
Reuters isn’t alone in seeing the potential for technology to augmented the work of reporters. Francesco Marconi, former manager of strategy and development at the Associated Press, wrote for us in 2016 that artificial intelligence “will help journalists do more investigative work by analyzing massive sets of data and pointing to relationships not easily visible to even the most experienced reporter.” (Marconi recently joined The Wall Street Journal as head of R&D.) The AP, too, has experimented with automated reporting via a partnership with Automated Insights, which helped it produce more corporate earnings and baseball game stories — perhaps the two most iconic genres of robot-friendly stories, each being derived from a structured set of data (official earnings reports and box scores, respectively).Despite the many benefits pushed by the boosters of automation, many reporters still may find it hard to shake the central concern about robo-reporters: that the tech could make certain kinds of journalists less necessary. A 2016 Tow Center report stated that claim plainly: “Automated journalism will likely replace journalists who merely cover routine topics,” wrote researcher Andreas Graefe, who also presaged the creation of ideas like Lynx Insights: “In the future, human and automated journalism will likely become closely integrated and form a ‘man–machine marriage,'” he wrote.
That, said Cassidy, is a central goal for the Reuters experiment. “A machine is never going to find a great quote and have perfect judgment about a story. That’s what journalists bring to the table. The point of this is not to take anyone’s job.”