“We want to take part in shaping the digital society”

How do you achieve impact through your research findings? For instance, by publishing a paper in a leading academic journal, or speaking about it at a public event or in the media. But the Data School's Mirko Schäfer and Iris Muis do not find the time to publish as often as they would like. And yet, their research is having significant impact, with their work used in e.g. Austria, Sweden, Finland and Greece.

Iris Muis and Mirko Schäfer, along with associate professor Karin van Es, are the figureheads of the Data School, an education and research platform exploring the impact of digitisation on society. Schäfer is an associate professor at the Science Faculty as well as co-founder of the Data School. Team manager Muis was nominated Responsible AI Leader by Women in AI Netherlands. Schäfer just returned from Finland, where the Data School team worked to raise awareness among scientists and government officials on the ethical aspects of big data projects. They did so with the Data Ethics Decision Aid (DEDA), a toolkit used to help officials by guiding them through each step of the decision-making process.

The Data School developed DEDA in 2016. Previously, the education and research platform had been conducting research on public debates online but noticed a lack of guidelines for responsible data practices. Working with data analysts at the City of Utrecht, who were grappling with the same issue, they created the tool to chart ethical issues in data projects.

‘Goose game’ for data ethics

“You can best compare DEDA to the Game of the Goose,” Data School team manager Muis explains. “You simply lay the poster on the table and the team working on a particular innovation - such as an algorithm, the deployment of drones or the creation of a dashboard - circles the poster and discusses the ethical pitfalls together and how to overcome those.”

DEDA has been used by several Dutch municipalities since 2016: Muis has presented at least two hundred workshops and lectures in more than 50 different municipalities. The Dutch government has included DEDA in its Ethical Innovation Toolbox. DEDA also gained prominence abroad: not only the Finnish government has had DEDA translated, in fact, the poster is currently available in English, German, Swedish and Greek.

DEDA enables people to learn more about digital ethics

Schäfer: “The beauty of DEDA is that not only does it provide a structured process for discussing and adapting the details of specific data projects, but it also provides the bonus of enabling people to learn more about digital ethics in general. Policymakers can identify shortcomings in their organisation and respond accordingly, by e.g. appointing an ethics officer.”

Europe comes knocking

DEDA is not the Data School’s only flagship project. In 2020, the Data School responded to a call from the Home Office. “This was during the aftermath of the so-called child benefit scandal in the Netherlands, which shone a real spotlight on potential adverse consequences of the use of algorithms and AI in government agencies,” explains Muis. “At that time, we developed the Fundamental Rights and Algorithms Impact Assessment (FRAIA). This is a comprehensive questionnaire allowing government officials to weigh any positive effects of an algorithm against its possible negative impact on human rights.”

Schäfer: “Most impact assessment tools look at the models themselves: is the model biased? Are the indicators and the data accurate? And while that is all relevant and important, it ignores all contexts. FRAIA, on the other hand, does consider this. The questionnaire covers various stages: what is the reason you want an algorithm in the first place? Then, what exactly should it do for it to work? What do you plan to do with the output? And finally, if there are any fundamental rights affected. And if so, is that acceptable? Can you justify it, to the public, to the media, to the courts? Would it be preferable to modify the design, or even not to use the algorithm altogether?”

It is important to highlight the variety in what can be seen as scientific impact

FRAIA has also taken off in a big way: the Dutch parliament has passed several resolutions to make FRAIA mandatory for all public-sector algorithms. In addition, the European AI Act included a Fundamental Rights Impact Assessment for the application of high-risk algorithms.

“Last year, Mirko and I discussed our experiences with FRAIA at the European Parliament and we are seeing interest from other European countries, which are naturally closely following developments around the AI Act.” For example, Norway's Digital Ombudsman used FRAIA as input for their own, Norwegian guide against discrimination in technology, and a Finnish official is coming to do a placement at the Data School.

Scientific insights

The beauty of DEDA and FRAIA is that not only are they tools for policymakers to evaluate their data projects, but they are also a method for researchers to gain insights into organisations' data & AI ambitions, their operational capacities and how they consider ethical issues. Muis: “We observe how participants think about technology. Do they see it as the answer to all problems or do they view it more as support? Do they only address privacy, or do they consider any other ethical aspects that might be important? How do they deal with human rights implications?”

To date, these insights have yielded two scientific papers as well as several professional publications and reports. Schäfer: “And because they are now also widely used by our academic colleagues in Sweden, Finland, Estonia, Germany and Greece, we hope one day to be able to outline a comparative analysis of local data practices in Europe, which does not exist yet. We don't know how different authorities deal with this.”

Dropping a paper

According to Schäfer and Muis, the Data School's work is public engagement par excellence, with public engagement being one of the pillars of Utrecht University's Open Science Programme. Schäfer: “We identify problems in the sector and work towards applicable solutions transparently. At the same time, this also helps us with our own academic research, but that is not where we start. We invest a lot of time in our network to gain insight into questions and urgent issues in the social sectors. As researchers we are looking for ways into these organisations because we want to study the use of algorithms from up close, and offer a contribution with real-life impact: we don’t just want to drop a paper at the base of the ivory tower. Our peer-reviewed paper on DEDA may not be the most cited, but it is one of that journal’s most downloaded papers overall. That could indicate that it is being used on the ground and I think that is much more important.”

If you want to achieve an impact, then a trickle down from universities to society will not do

Consequently, the two researchers are keen to push for a broader definition of “impact” at university level. Muis: “In academic circles, it is important to highlight the variety in what can be seen as scientific impact. The way we work is a different way of doing science, a different way of thinking. That also has impact, because it provides enormous added value for all concerned.”

Schäfer: “Our way of doing research and teaching facilitates effective, mutual transfer of knowledge. There is a tendency these days among many university admins to simply appoint an impact officer to re-establish contact with society. But often it just doesn't work that way. The corona pandemic, war, climate change, the digital transition, these are all topics assessed and commented upon in society today, often in real time. That requires an approach to research that is also firmly rooted in society. If you really want to achieve an impact, then a trickle down from universities to society will fall short. It is our ambition not only to study society, but also to help shape it.”