Artificial active

How can artificial intelligence help increase audience engagement with news?

Credit: Ketut Subiyanto from Pexels

Successful solutions are never about trying to create a new product, but about correctly defining the problem we are trying to solve.

That was the approach of more than 20 news outlets around the world who worked together last year to solve some of their problems, using artificial intelligence (AI) technologies. The Collab Challenges – as these experiments are known – are organized by JournalismAI, a project of Polis, the journalism think tank at the London School of Economics and Political Science (LSE) and supported by Google News Initiatives.

Many media outlets were looking for new ways to reach audiences they weren’t serving well or to increase the value of their journalism and retain their readers.

Reaching young readers

One of the participants was Ketil Moland Olsen, senior project manager at Media City Bergen, who was looking for better ways to reach young readers who rarely read the news.

Thanks to the data collected from several newsrooms, he and his team have realized that 18-30 year olds often have difficulty understanding the news because they do not know the background of the stories or even their protagonists, who are generally more aged. . However, they have the same appetite for information as older generations and want learn more journalism.

After testing news content with a sample of young readers, the team set about creating a tool that would identify difficult topics and unusual words that journalists often take for granted. They then imagined a “micro fact box”, a small explanatory box that unfolds when the Internet user clicks on these words that the tool highlights in the article. This way, the reader has the choice to find out more about the topic or a person if they wish, or to continue reading undisturbed.

Prototype of Micro Fact Box. Info boxes between paragraphs become visible when the user clicks on a highlighted keyword or phrase in the text. Courtesy of Media City Bergen

At the moment the text is taken from Wikipedia and is not editable by newsrooms, but the team is looking to change this in the future.

Find the right balance

There’s a fine line between being helpful and being intrusive or coming across as condescending. So the question is – how much background is too much?

There are several ways the team is looking to define the number of unknown words that should have a micro fact box. One way is to identify the user as a young or new subscriber and present them with more definitions, assuming they have less knowledge. Conversely, older, long-time subscribers may receive fewer keywords.

But other scenarios are explored, for example, building an algorithm that learns how often a user views facts, then presents them with more or less explanation in the future. Another possibility is to offer new subscribers a quiz where they can test their knowledge and then create a profile for them.

All of this comes with ethical issues around user tracking, GDPR, and testing people’s knowledge. On the other hand, the tool needs user data to work properly.

This could be solved by giving the user a choice of how many explanations they want, much like when you set your game profile as easy, medium, or hard.

But the challenge with the user-led approach, Olsen said, is that people tend to overestimate their knowledge of many topics or aren’t completely honest with the media or themselves.

Understanding climate science

A similar approach was attempted by Meik Bittkowski, head of R&D at Science Media Center Germany, and Marco Lehner, technical product developer at the public service broadcaster. Bayerischer Rundfunk.

They realized that younger, less educated audiences had a harder time understanding stories about climate change, and journalists were often tired of re-explaining basic concepts.

So when posting climate stories, the team experimented with a fact box that gives the user a factual explanation of climate science, with a traceable reference to the source, whether a climate expert or a study published in a scientific journal. They hope that this transparency will help increase public confidence in scientific and journalistic sources.

Unlike the micro fact box, these explanations are written by journalists and can be modified ad hoc. This is important when dealing with science, as data may require frequent fact-checking and updating. The ready-to-use boxes can also be used in multiple stories and can be useful when reporters from different bureaus need to cover climate change as well as their usual pace, be it news, economics or sports.

Reader loyalty

Another medium seeking to boost audience engagement was South China Morning Post (SCMP) where Assistant Digital Editor Shea Driscoll wanted to find a better way to recommend relevant content to keep readers hooked longer.

Almost every news site has a version of the “Read more” section filled with content that editors choose based on their editorial judgment. The catch is that most humans think of a small number of articles that can be linked to, rather than crawling the whole website.

But what does better does it mean when it comes to content recommendation? Is it something that will educate the reader in more depth on the same subject, or an entirely different article that will broaden their horizons?

To answer this question, Driscoll’s team analyzed the performance of content on the SCMP website and mapped out which stories people read after finishing the first one. They’ve come up with an algorithm that works much like a shopping website recommending additional items based on “people who bought this also bought that.” It could be from any other section of the website and not related to the first story at all.

The results were surprisingly encouraging. When testing the tool, the team found that readers who saw content recommended by the algorithm were 75% more likely to return for more. Even better – those who read one article and recommended another from a different section were twice as likely to read an additional article from a third section of the website, which significantly boosted content consumption.

How the recommendation was positioned also mattered. When it appeared in the middle of the article and the user needed to actively click on it, they were more likely to read the article. However, if the article started after the end of the first (infinite scroll), the user was more likely to give up.

Driscoll said this feature could play an important role in SCMP’s subscription strategy because reader loyalty and the amount of content they read are important factors in retaining subscribers.

Free daily newsletter

If you like our news and feature articles, you can sign up to receive our free daily (Monday-Friday) email newsletter (mobile friendly).