News

AI in local journalism

10 Feb 2022

Artificial intelligence is poised to support local journalism in the future. In this interview, Professor Neil Thurman and Dr. Bartosz Wilczek discuss their new research project.

A newspaper roller during printing | © IMAGO / Westend61

Local journalism is facing many challenges. For instance, declining print sales are making it difficult for local news providers to stay afloat. AI-based applications could offer some relief by carrying out research, writing articles, and optimizing distribution.

An international project comprising experts from Munich, Leipzig, the Netherlands, and Norway is researching and developing just such applications. Interdisciplinary by design, the project brings together researchers from the fields of computational communication, computational linguistics, media management and economics, as well as AI law and ethics. At LMU, the project is coordinated by Professor Neil Thurman and Dr. Bartosz Wilczek and is based at the Institute for Communication Studies and Media Research.The Volkswagen Foundation is funding the project with EUR 1,412,000 over the next three years.

Local newsrooms are in crisis, it’s often claimed. But what exactly does this mean?

Neil Thurman: Local news outlets are facing challenging circumstances. They are having to deal with declining print revenues, find better ways to monetize their online content, react faster and compete with larger newspapers. As such, they need support to become more efficient.

There is already evidence of how AI can help local journalism. In the UK, for instance, there is a news agency that provides automated articles to local journalism clients. They are sold for about one twentieth of the amount it would cost if a local journalist had to write the individual story him- or herself.

We’re looking forward to working with local news organizations in Germany to realize some of that potential with them and then share the lessons learned.

Reducing costs while working in a journalistically responsible manner — that sounds like a big challenge. How do you address the ethical aspects?

Neil Thurman: “Responsibility” is the key word here. While of course we want to increase the efficiency of processes and work more cost effectively, it’s also hugely important to us that we live up to our responsibility in meeting journalistic values and standards. This is where an ethical approach comes into play. We can design and program AI-based applications so that journalistic values are built into them.

Incidentally, the definition of AI is important in this context: Our project is focusing on Artificial Intelligence, but we’re also talking about more simple ways of automating processes, which perhaps don’t always meet a strict definition of AI.

Can you give any concrete examples of how AI journalism would work?

Neil Thurman: Well, there are different aspects that seem very important to us. One of them is research. Finding newsworthy stories is an area with much potential. Social media offers a lot of potentially newsworthy information. However, because of the sheer volume of posts on social media, it is difficult for journalists to screen them manually.

There are already tools in operation that can automate the process of finding information on social media. They can search for events or discussions that have not even been reported yet by the mainstream media, and alert journalists to leads.

Another potential area is the distribution and personalization of news. Even though personalization of news goes back around 30 years, there is still a lot of potential and room for improvement, such as making sure that readers are exposed to news of sufficient diversity. You might not want to just show people what they requested, but also news that they might need to know. During the COVID-19 pandemic, for instance, this aspect gained a lot of relevance.

It’s actually difficult for people to predict what news they’re interested in. We’re not very good judges of our own interests when it comes to news, because they change so fast. If you were setting up a profile of news that interested you in 2019, most people would never have chosen to get news on COVID-19, for instance. This is why so-called collaborative filtering systems are useful. They match the interests of people who appear to be similar in various ways, which doesn’t require people to describe their interests in exact detail.

Moreover, for local journalism in particular, subscriptions are a big part of the financial model. Even here, AI and automated processes can and will help the news publishers — by looking at reading patterns, picking up other cues, and trying to incentivize readers to keep their subscriptions or to subscribe.

Artificial intelligence that writes a news article: Is that a realistic prospect?

Neil Thurman: Well, it’s realistic in that templates can be created that can produce stories automatically from data sets.

Although the stories are produced automatically, they’re done within a framework that the journalists and technologists have designed. For example, the data comes in and, depending on the information, the machine will write a certain sentence or a certain paragraph. It’s not making its own decisions about how to do this; it’s merely following instructions. The parameters are set by the journalist, even though the computer is producing the text in an automated way, following the instructions.

How are local news editors and journalists responding to the new technological opportunities?

Bartosz Wilczek: In the planning phase of our project, we carried out two studies. The findings show where local news organizations see the biggest future potential of AI along the news value chain, and they indicate conditions that might facilitate the adoption of AI in local journalism. Drawing on the findings, we developed our research project.

We already have a number of local news organizations that are interested in collaborating with us on the project. This clearly shows us that the interest is there.

We’ve already touched on the English and German media landscapes: Are there differences between them that would influence the use of AI?

Neil Thurman: Introducing AI to the media system is probably easier with English-language media, because many of the tools have been developed with the English language in mind.

Although there are technology service providers in Germany who have systems that operate in the German language, there is a need to extend the benefits of automation into non-English language settings — something that is already starting to happen.

How did your project start?

Neil Thurman: I’ve had a Volkswagen Foundation Freigeist Fellowship since 2015, on the topic of AI and automation in news.

This new project, which is also funded by the Volkswagen Foundation, is different because it focuses on the local news market. It’s also more focused on developing tools as well as doing traditional social science research. We’re really looking forward to working with the local news organizations to develop the tools, and to share what we learn from that process with the local news community more widely.

Bartosz Wilczek: Our research group consists of amazing international and interdisciplinary scholars.

We will not only develop AI tools together with local news organizations, we will also investigate intra-organizational, inter-organizational and societal conditions to facilitate the adoption of AI in local journalism. Thereby, our project will have a long-lasting impact.

What are you looking for?