Data Science: Discovery where disciplines intersect

19 Aug 2021

Newly appointed at LMU Munich: Frauke Kreuter investigates data with a focus on quality.

Prof. Dr. Frauke Kreuter

Positions her research at the point of intersection between statistics, informatics and the social sciences: Frauke Kreuter. | © LMU

In the formation of a political will, digitalization and social media are a mixed blessing. They enable stakeholder groups to forge networks more quickly. They facilitate dialogue. They serve as a coordination platform for movements such as “Fridays for Future,” and can give even quieter voices a chance to be heard. Yet they can also amplify discrimination, lead to destabilization, and advance the cause of antidemocratic forces. The storming of the US Capitol on 6 January of this year is but one example of what mobilization against democracy can achieve via the Internet.

All digital processes are rooted in data. Masses of data. And in turn this data lays the foundation for algorithms that, for example, determine what content and offerings a person is shown most prominently in the digital space.


Digital media: “We often don’t have the right data”

Read more

Professor Frauke Kreuter, the Chair of Statistics and Data Science in Social Sciences and the Humanities at LMU’s Department of Statistics, has a keen interest in the digital process that drives these algorithms. “We investigate data with a particular focus on quality,” Kreuter says. She and her team analyze how data is captured, what groups need it and for what purposes, and how many errors it contains, for example; but also what data are missing and how they might be used going forward.

However, accessing these data is not always easy for researchers. “Right now, we often don’t have the right data at our disposal to really dig deeply and find out what is happening in digital media,” Kreuter explains. “Empirical studies show that attempts have indeed been made to shift public opinion - but measuring the precise effects of these efforts is difficult.” Accordingly, she calls on big tech corporations such as Facebook and Google to simplify access to data for the purposes of research.

‘Fair’ algorithms


Big data are no substitute for personal input in surveys

Read more

In this context, Frauke Kreuter also studies the relationship between algorithms and fairness. Algorithms are ‘trained’ with existing data, which means they can perpetuate long-held stereotypes and biases. “I like to explain the point with the example of image search on Google,” she says. “A couple of years ago, if you typed in ‘university professor’, you would have seen only male professors, with a good portion of African-American professors in the search results but no females. Then in 2019 within a month the algorithm changed and now shows almost about equal numbers of women and men.” Kreuter sees such sudden changes as indicative of how difficult it is to get a proper picture of society if all we do is rely on digital data traces. In many applications when algorithms are trained on historic data, we tend to get a trailing picture of society - like what we see looking at the rearview mirror while driving. However, the previous Google algorithm was not entirely wrong, she notes, because – in Germany at least – there are still more male than female professors.

Frauke Kreuter’s appointment at LMU will further reduce this disparity, not only because she is a female professor but because she brings social science background into the department of mathematics, informatics, and statistics. Kreuter studied sociology at the University of Mannheim. Immediately after earning her doctorate in empirical methods in Konstanz in 2001, she traveled to the US as a postdoctoral student to the statistics department at the University of California, Los Angeles (UCLA). She then moved to the University of Maryland in College Park (UMD), first as faculty member and eventually director of the Joint Program in Survey Methodology, a program jointly run by UMD, the University of Michigan, and the data collector Westat.

One foot in LMU

“Originally, I intended to spend six months at most in the USA. But I liked it so much that I stayed,” she recalls. Above all, she was thrilled by the flat hierarchies and the chance to have a say in lots of official bodies. Six months turned into twelve years – a period during which the data scientist first established a relationship with LMU. Through a cooperation agreement between the LMU and the Institute for Employment Research (IAB) in Nuremberg, she was hired at LMU and loaned to the IAB to head and grow the Kompetenzzentrum Empirische Methoden (Empirical Methods Competence Center), a unit that advises and supports IAB researchers on data collection and analytics, as well as advancing research into the exploration of new data sources from which to capture unemployment and occupational activity. Here again, data quality was an important topic. “The work involved issues such as how you structure surveys in order to obtain a solid data basis and be able to provide the government with sound advice,” Kreuter explains. Alongside her research work in Munich, the professor continued her activities in Maryland. To work simultaneously on two continents: “I was relying on Zoom meetings before Germany had ever thought about them,” she laughs. Kreuter saw both activities as a good way to “build bridges” and promote scientific dialogue – something she plans to continue in her work at LMU.

After her time in Munich, she accepted a chair in Mannheim, where she continued her work at both Maryland and the IAB. In partnership with both universities and with the support of Germany’s Federal Ministry of Education and Research (BMBF), she set up the first fully online master program in survey and data science. Following a year in California in research residencies at Facebook, Stanford and UC Berkeley, she returned to LMU in 2020.

Frauke Kreuter positions her research at the point of intersection between statistics, informatics and the social sciences: “At both Maryland and Mannheim, I have established data science centers that pool expertise from all relevant disciplines,” she says, fully intending to continue the same approach in her work in Munich. “Initial talks have been very promising!” she adds with a smile.

What are you looking for?