Between journalism, data science and mathematics: a job with many facets.

How Andreas supports journalists in their work with data and software applications.

What is your name and what do you do at ZEIT ONLINE?

My name is Andreas Loos and I work as a Senior Data Scientist in the editorial department of ZEIT ONLINE.

Why did you decide to join ZEIT ONLINE?

I've been a fan of print ZEIT since I was a teenager and became a fan of ZEIT ONLINE when it became available. For many years I worked as a science journalist and as a mathematician. At some point, I thought I might be able to combine the two at ZEIT ONLINE and applied on my own initiative. It worked out - but in a completely different way than I had imagined at the time.

What are you and your team currently working on?

Our work is extremely diverse and agile. By that I don't mean the "agile" from process management, but real agility: literally new tasks and challenges every day, new scientific methods, fresh ideas. We are constantly on the move, somewhere at the interface between software development, data engineering and data science, between natural language processing, statistics and data journalism. Sometimes there's also a bit of mathematical optimisation or graph theory involved.

We use them to build applications that help online editors work better and faster and to create journalistic projects that sometimes involve a lot of mathematics or computer science.

What has been the most important project for you and your team so far?

In the last few years at ZEIT ONLINE, I have worked on a lot of very exciting projects. The most important to me personally were always the tasks in which journalism, social commitment and mathematics came together very closely. Examples are the "49" project or the "My Country Talks" project, which emerged from "Deutschland spricht" (“Germany Talks”). In the "49" we selected 49 people from 30,000 applicants in such a way that they reproduced the respective distribution in the German federal population in about 70 factors (income, size of place of residence, age, education, etc.). At the time, we formulated this as an integer linear optimisation problem. The 49 people then spent months discussing very different topics with each other and brought very interesting questions to the editors.

The project "My Country talks", on the other hand, is essentially a graph-theoretical matching problem: we try to bring together thousands of people with as many different political opinions as possible in Germany or Europe in pairs so that they can discuss political and social issues. Soon we will be networking hundreds of thousands, maybe even millions of people worldwide in this way at "The World Talks" - I am very curious to see how many will participate.

What topics are you already looking forward to?

We are currently in the process of developing new algorithms, especially for clustering and classification problems, which take better account of the structures of the data. We then want to apply these algorithms to different data, for example to our corpus of articles or to social networks, as they repeatedly play a big role in the data of the Investigative Team, for example. That's interesting from a research perspective, but also in terms of content.

The rapid development of large language models (LLMs) such as GPT or Llama is also very exciting at the moment. The big question is whether and how such models can be used in journalism. Suddenly, there are not only technical but also interesting ethical questions on the table that were previously discussed in smaller circles, among data journalists or in data science: What is the bias of the training data and how do we deal with the fact that we often cannot control it? Can we measure the quality of the models or the results at all? And how can we guarantee quality in the end?

What do you particularly appreciate about working at ZEIT ONLINE and in the field of data science?

On the one hand, ZEIT ONLINE is very open and flexible when it comes to fresh ideas, and on the other hand, it has a very lively culture of discussion. It's easy to find interested people to gather ideas, to get projects off the ground together - or even to discuss the value of solutions. Critical questions quickly arise. That can be very helpful. Journalists are particularly good at asking critical questions.

Back to overview


Senior Data Scientist at ZEIT ONLINE

"We are constantly on the move, somewhere at the interface between software development, data engineering and data science, between natural language processing, statistics and data journalism."