#MILGlobalUnderstanding

#RethinkingHumanHorizon

Entrevistas

Olle Zachrison: “We have to help people navigate in the flood of information that meet them”

Fotografía cedida por Olle Zachrison.

Versión en español

The Swedish public radio, Sveriges Radio (SR), has created an editorial algorithm that promotes civic values. The local newsrooms of the public radio began to use this algorithm in September and it has been gradually implemented in 26 local newsrooms. According to Olle Zachrison, Head of Digital News Development, this system is helping both the workers and the public.

The media continues to reinvigorate itself while further adapting to technology. Artificial intelligence systems as well as automated mechanisms that help journalists are being brought in progressively. Sveriges Radio has incorporated an editorial algorithm that will allow its editors to improve the time and quality of their work. Olle Zachrison explains how this system works and its implementation in public radio.

The use of an editorial algorithm in the Swedish public radio is an innovative approach. Tell us about your assessment to this experience; to what extent has it been successful?

It has been very successful in the way that it’s a system that we have been able to introduce pretty smoothly in the newsrooms where it has been integrated in our editorial processes. I think a key to success has been the focus on journalistic quality, not on the technical applications. When we rate our news items, the main purpose is to evoke a journalistic discussion about the importance and quality of that individual piece. The fact that these “news values” also feed the editorial algorithm is important but it’s not the main point. I think this approach has helped us internally.

How far has the system with which the algorithm works met its objectives?

It has met the objectives insofar that we can now save substantial amounts of time when we let the algorithm sort the running order on our local websites and in our local news playlists. The editing has also become less arbitrary when everybody is using a standardized system, in all our local newsrooms. The algorithm has been trimmed in, and it does do its job by giving the audience a mix of news which makes sense and feels close to what a completely manual system would produce. But, there are a lot of challenges ahead, like using the algorithm for our national websites. This is more complex as the information flow is much greater and news items come in from so many different SR newsrooms.

How can you guarantee that the assessment generated by the algorithm regarding the news being trustworthy and reliable is accurate?

This is guaranteed by letting editors do a careful editorial assessment of every news clip. The algorithm doesn’t work completely independently but has to be fed by the “news values” that our experienced editors attach to every news piece. Nothing is produced by the algorithm that hasn’t passed human scrutiny. Even if we now let the algorithm sort our local listings, that mix is always supervised by human editors that can change parameters if they think the output looks strange or editorially flawed.

You have previously stated that you would like to create and implement customised AI-supported searches. Personalization engages and helps users, but at the same time they become exposed only to related content which appears to contradict with the mission of public service media and with the idea of media pluralism, how far do you agree and how can a balance be achieved here?

I think that risk certainly exists on platforms like Facebook where people meet a small selection of content designed to maximize engagement and where a lot of content is not trustworthy in the journalistic sense. In the SR system, all news content is carefully assessed and approved by editors. Moreover, we will always feed all our visitors what we deem is the most important news of the day. So the risk of being fed just one type of news or perspectives will be very low. We foresee that we will let the algorithm curate hundreds, maybe thousands, of news playlists. One person maybe want to have news from two of our local stations, plus the most important news in Arabic but only sport if it’s really important. All those combinations are impossible to curate manually. But we set up the system in a way that ensures that the mix we produce will always be in line with our mission. We also think that introducing a special set of public service values safeguards this. Stories with unique perspectives, with outside reporting from communities and with strong audio storytelling have a special prominence in the system.

How far are algorithms considered to be a real solution to enhance journalistic quality in your opinion? 

The algorithms are not an end in themselves. But in today’s media landscape, the audience expects so much more of our services. The standard in the household digital services is that everybody is not fed exactly the same content. We have to help people navigate in the flood of information that meet them, also on our own platforms. Advance programming and algorithms is just an aid to do this job better, to become more relevant. We know our content and our different programs so much better than a lot of our visitors, so if we can use computational power to help people discover news and shows that they are surely interested in but don’t know about, that is a positive thing. Importantly though, I think we always have to go back to the question: “How can we increase the quality of our news outlet?” And if we can build a technical system that can actually provoke a more fruitful debate about the journalistic content we produce, then I think we can kill two birds with one stone. Working on our “public service algorithm”, the mere discussion about what that term means has proved very inspiring.