Should editors burn their fingers on algorithms?

Do you only see news about Britney Spears popping up on a newspaper’s website? Then there’s probably an algorithm gone haywire. How do you safeguard an editor’s journalistic mission if automated algorithms determine what you, the reader, get to see? Postdoc Max van Drunen investigated so-called "news personalization.

Computerization

’We have about 300 years of experience with the analogous way of making editorial decisions: journalists determine what goes on the front page of a paper newspaper. Now we are dealing with ’’news personalization,’’ says Van Drunen. So what are we talking about? ’’An article passes by on one person’s screen, but not on another’s. Those decisions are made automatically based on what people click on and how long they look at something. So an algorithm makes editorial decisions.’

Implementing news personalization

’Almost everyone is experimenting with it,’ Van Drunen saw in his doctoral research. The Volkskrant, for example, is already pretty far with the implementation of news personalization, while the NRC had actually renounced it. ’The NRC simply said: we are not into this scary algorithmic mess. They have since partly gone back on that. But the NOS, for example, is also cautious about it. Many media companies are struggling with it. How do you promote the company’s values if you use news personalization? They are afraid of reputational damage if that goes wrong.’

Impact of algorithms

It goes wrong, according to Van Drunen, when algorithms start determining content: people read a lot about the British royal family, so we only write about that. Or if journalists write about all sorts of things, but a reader only sees articles about the royal family because of news personalization. ’We know enough stories by now about filter bubbles on social media and polarization. So you have to measure the impact of algorithms properly.’

Many media companies struggle with it. They are afraid of reputational damage if it goes wrong

’Tailored news’ also has advantages. It makes money if readers stay longer on a site because they see content that appeals to them. The goal is also to provide readers with content they are actually interested in. The big question is how to reconcile that with the values a company wants to convey. How informative is a piece, for example? Or how diverse is the news that people see? The latter is in several aspects: which sources are used and which political parties or perspectives are covered in a piece? Such things are as yet difficult to measure with an algorithm,’ Van Drunen explains.

Ethical code

According to Van Drunen, the next step for media companies is to work out a code of ethics that an algorithm must comply with. ’That is currently still lacking. Many norms and values are about the quality of articles. That is very important in a media system in which not much content is produced. But now there is so much content that an important question is which content is actually shown to people. The editorial values in that area are not yet well developed.

Max van Drunen is a postdoctoral fellow at the Institute for Information Law (IViR). He investigates the regulation of political microtargeting as part of an interdisciplinary research project combining communication science, philosophy, computer science and law. He also continues to explore how editorial responsibility of platforms and media is shaped by the automation of editorial decision-making. His dissertation was conducted as part of the Personalized Communications project under the supervision of Natali Helberger.

State influence is sensitive

Are there many legal restrictions with news personalization? ’Not that much. The law has a difficult position in this. State influence in editorial decisions is very sensitive.’ The current ground rules around news personalization are laid down in the AVG and the European Digital Services Act (DSA). ’But if I look at this with a traditional media freedom lens, I would scream blue murder. The rules in the DSA are vague and give governments a lot of room to apply them. Still, according to the researcher, there is no immediate cause for concern because the independence of media companies is also well guaranteed at the European level.

Van Drunen is more concerned about creating a well-functioning algorithm. It takes a lot of time and money to develop that. ’It requires close cooperation between journalists and engineers, otherwise it will go wrong. Right now, only the rich media companies, which might accept money from Google or Facebook, can further develop news personalization. On the other hand: if the readership of the Volkskrant has exploded in ten years, the NOS will be scratching its head.’