The newest writer at Fædrelandsvennen (FVN) in southern Norway is a hard worker who doesn’t ask for coffee breaks or weekends off. And that lack of leisure time doesn’t hurt output, which is expected to be around 4,200 stories a year.
If that sounds like an inhuman amount of productivity, that’s because this writer is a robot. And it’s part of a growing trend in Nordic countries to use AI and robots to gather information for stories.
FVN began using a “robot journalist” in February, said project lead Mads Ommundsen. The robot reports on real estate transactions, such as which homes are bought and sold in certain areas, the selling price of each home, and more.
“We have seen that other newspapers we work closely with have had a significant improvement in the number of articles concerning real estate after the introduction of automated articles,” Ommundsen said. “We have a long tradition for publishing real estate sales; with automated articles, we were able to present this information in a more streamlined way online.”
FVN “introduced” its robot journalist in an article explaining how the process works and showcasing some of the articles it had already written. The robot is a service provided by a Swedish company, United Robots, which also works with publishers Bergens Tidende and Aftenbladet.
“What got me interested in the opportunity are the articles delivered by the robot to Bergens Tidende,” Ommundsen said. “We have seen how they have built up a database of articles using only a text robot.”
AI for real estate
The robot gathers information from public land registers and also gathers property data and statistics including details about the sale and the property itself. Then it calculates the price per square meter and compares it with other nearby home sales. Each article also contains Google images from both the street level and a satellite view.
The greatest challenges for FVN were deciding what information should be included in each story and how to divide up coverage areas geographically. Ommundsen said they also had to decide the best way to present the articles on the front page, and ultimately settled on having a fixed position for the robot-generated content. The new “employee” allows FVN to dramatically boost the kind of attention it gave to its real estate coverage.
“Before the robot, we almost exclusively wrote stories on real estate purchases if it was a large sum or famous people who bought or sold real estate,” Ommundsen said. “We now publish all purchases and sales. So where we may have had around one article a month on average before, we now have around 350 articles a month.”
Robots vs. newsroom
While the robot generates articles quickly and at low cost, Ommundsen is quick to point out this is a supplement to live journalists, not a replacement.
“The robot can never do as good a job as a journalist,” he said. “The advantage is that we can have this type of material without stealing resources from more important work. Real estate and real estate prices are popular [content]. This is in addition to traditional news.”
Although the initiative is in its early stages, Ommundsen already sees an uptick in readership for its real estate coverage. This will open the door for more applications in other sections in the future, he believes.
“A lot of exciting things are being done in this field. If the data eventually becomes good enough, it could, for example, be an opportunity to look at robot journalism in areas such as sports and traffic incidents.”