At Schibsted, we are convinced Artificial Intelligence (AI) has immense potential to strengthen journalism and news practices. If AI is used in good and responsible ways, we believe these new tools can, for example, help prevent filter bubbles, contribute to sustainable financing of media operations, and increase diversity in news representation.

An example of the latter can be found at Bergens Tidende (BT), where AI is paired with a low-tech research community — proving the value of combining human and machine expertise.

Insights to improve journalism

BT’s small and agile audience engagement team is on a mission to improve the creation and reach of editorial content. Based on Norway’s southwestern coast, the team delivers quantitative and qualitative insights to enable editors and journalists to showcase journalism in new ways and make smart choices in their work.

To better serve its broad group of readers with relevant and engaging content, BT strives for more diversity in its news coverage. A key aspect of this effort is to make BT more relevant to women across various ages. Here, we will look at two examples related to this.

Using computer vision to understand newsroom biases

In an effort to get more female readers (and, ultimately, subscribers), BT is using the AI concept of computer vision to better understand who they are telling stories about. Specifically, BT is employing it to systematically uncover what people the imagery on its Web site depicts.

By estimating the age and gender of faces used in an article’s imagery, BT’s application of computer vision enables insights into how its news coverage relates to the demographics of its audience. Adrian Oesch, a data scientist on the audience engagement team, has been leading the work.

“In general, I think it’s super interesting to explore how technology can help improve our understanding of newsroom processes. In this particular case, we saw computer vision as a valuable method because it’s essentially automating a repetitive task we could have spent human resources on,” Oesch said.

This following chart shows the average share of female faces in images and the average share of female readers for the 100 most popular topics in BT. The chart shows the share of female readers increases as the share of images with females does.

Young women give their input

Image recognition is not the only tool employed to shift toward more representative news coverage. For example, BT has studied what topics and themes are most read by men or women, elderly versus younger people, and local versus national readers. These are just a few of the efforts made to diversify the content produced.

Hanne Louise Åkernes, deputy news lead at BT, is running an effort called Project Silje together with political commentator Gerd Tjeldflåt. Project Silje could be seen as a more low-tech equivalent to the image recognition solution. By creating a physical and digital arena for young women to meet and discuss news topics relevant to them, Åkernes and colleagues are extracting learnings about how to create editorial content that makes young women connect more with what is written in BT.

“At times I worry that media organisations forget to listen to what readers really care about. In order to attract a new generation of news readers, we need to open up for readers to tell us what kind of content they want to consume,” Åkernes said.

Through an online survey, a Facebook group, and physical discussion events, Åkernes and her BT colleagues have so far given about 5,500 women aged 25-40 the opportunity to tell them what they are interested in learning about through the news.

“This new way of interacting with our readers gives us an opportunity to dig into topics that really matter to our target audiences. So far, the content produced through this process has performed very well in terms of clicks and conversions,” Åkernes explained.

“Team Silje” and COVID-19

Team Silje has helped the BT team create editorial content during the coronavirus crisis. Åkernes and colleagues reached out to the community with an open question: What do you need to know in your daily lives in this new reality?

A digital workshop with members of the community resulted in a news story about how coronavirus has changed our lives. Activity in the Facebook group led to many stories about holidays during the pandemic, personal finances, and family health, for example. The stories coming out of the collaboration with the Silje community led to increased subscription sales and clicks on the BT site, performing particularly well among female readers. Two of the stories were even among the most read and sold during the initial coronavirus lockdown.

Exploring what tools to test next

The combination of quantitative and qualitative user insights, paired with an organisation that loves testing new solutions, makes BT a hotbed for new journalistic solutions. To be able to work for more diverse representation in the media, BT is now exploring what other tools might provide journalists and editors with the opportunity to make more informed choices.

Banner image image courtesy of Wolfmann | Wikicommons and Pete Linforth from Pixabay.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .