“You don’t have to become a developer or a hacker, but you have to have an understanding of how algorithms work – and why they do what they do. You need to understand why the rules are as they are. That is the key,” Johnsen said.
“This is something you need to understand: It’s going to affect your work-life one way or another,” he told participants at WAN-IFRA’s Media Leaders Summit Middle East in Dubai.
Adresseavisen is both a traditional publisher as well as a thoroughly modern one. First published in 1767, and based in Trondheim, it is Norway’s oldest daily newspaper and part of Schibsted and Polaris Media.
The publisher’s newsroom has also been working with AI for many years, and has learned a number of lessons, which Johnsen, Adresseavisen’s Editor and Head of Editorial Development, was happy to share.
Adresseavisen primarily has used algorithms for personalisation and to increase the number of stories they could offer readers on their homepage, he said.
To provide some context, Johnsen noted that in Norway, unlike many other countries, about 80 percent of users actually come directly to the publisher’s homepage, rather than finding them through their content that shows up on search or social media.
“That means when we try to innovate, we usually do it on our homepage,” he said.
Creating the best possible homepage
Adresseavisen’s goal, therefore, is to create the best homepage they can.
Johnsen said when they started personalising, they began to consider whether the user was or was not a subscriber. Were they interested in hard news or sports? Did they prefer long reads or trending stories?
“We found that AI or algorithms worked best at the bottom of articles,” he said. “So when you were reading one story, we had a bigger chance of getting you to read one more and one more and one more.”
For example, he said many sports fanatics will read 10 stories about sports when they are on Adresseavisen’s website.
The publisher’s efforts focused on “trying to get you to be a subscriber, or get you to read one more story, depending on what kind of user were you were,” he said.
‘You need to have enough data’
Adresseavisen was basing this personalisationlargely through tags they had for stories: People who clicked on enough sports stories became a sports lover, and so on.
“If you are going to personalise news, you need to have enough data,” Johnsen said. “That means you have to have enough stories. If you are publishing 15 to 20 stories each day, you don’t have enough content to personalise. And you have to have enough users so you can build a group or you can build segments for the users.”
There are also certain kinds of stories that can be created largely by and through AI. For example, Adresseavisen is using AI to take data around area home sales (price, former owner, etc) to produce simple stories about local real estate sales.
While Adresseavisen has found many areas where AI can be useful, there are some others where it is off limits.
For example, the top of Adresseavisen’s homepage, and its lead stories are strictly controlled by editors.
Supported by AI, but controlled by human editors
“We don’t use algorithms on our top stories, and that’s just to be true to our mission,” Johnsen said. “You can do a lot of breaking news at the top and make sure that subscriber stuff and hard news can be shown by algorithms further down and it’s much more sophisticated.
“We can give subscribers and non-subscribers two different front pages depending on if we want you to become a subscriber or read as many stories as possible each time you are visiting us,” he said.
One of the most popular ways Adresseavisen is using AI today is to create stories based on the latest electricity prices in Norway, because these are fluctuating widely.
“It’s going up, and it’s so much every minute. People are going to our pages right now just to find out ‘Should I take a shower now, or should I wait two hours?’ Johnsen said.
This is exactly where AI can come in handy, he said, because it would not be worthwhile to have a journalist spend the amount of time required to continually update the figures.
Looking ahead, Johnsen said they plan to develop their own AI programmes. They are also looking into how AI can help them repurpose content across other platforms.
A word of caution
One thing Adresseavisen forbids its journalists and editors to do is to run content they’ve not yet published through third-party AI, Johnsen said.
“We have a rule in our newsroom that you never put unpublished material into ChatGPT,” he said. “The easy answer to that is because it goes into a database and you don’t know who is collecting your data. So that means if you have sources, or part of the story has not been published, someone else can be looking over your shoulder. Always remember that when you are using third-party.”
