Automated Decisions: Europe must speak with one voice

Automated decisions have become a part of many Europeans’ daily lives. Whether for job searches in Finland, healthcare functions in Italy, or the identification of neglected children in Denmark, such systems are coming into use in many EU countries – often for core public-administration functions. Will the EU be able to develop a common response on these issues?

In our new report “Automating Society – Taking Stock of Automated Decision-Making in the EU,” togehter with AlgorithmWatch we cite more than 60 case studies from around Europe in which automated systems are assisting or even independently making decisions (ADM systems). In doing so, we demonstrate for the first time that this is not a phenomenon limited to the United States or China. Over the course of the study, we review examples drawn from 12 EU countries, while additionally analyzing current legislation and the coordination of common standards at the European level.

As we further determine through the results of a representative public survey, nearly half of the European Union’s citizens do not know what an algorithm is. This despite the fact that a variety of ADM systems can be found in all member states examined, as well as at the European level. Moreover, a number of these are being deployed in sensitive areas of life. In Finland, for example, job applicants open their email boxes to companies. HR managers then decide which candidates are most promising on the basis of automatically created personality profiles. The Finnish data-protection ombudsman has expressed doubts as to whether this practice is legally sound. In the Italian healthcare system, an ADM system helps choose the most effective and efficient treatments for patients, with the additional aim of optimizing financing through the use of data-driven decisions. The Danish case study shows how three communities have developed a system that uses risk parameters for the early identification of children which are vulnerable to neglect.

More than just examples from the field

It is particularly striking that public administrations themselves use ADM systems in many EU countries. This makes it all the more urgent to engage in broad societal debate on a key question: How should the state address the possibilities presented by new technologies in order to ensure they benefit society? Citizens must know when automated decisions are being made, and for what purpose. And public agencies must be systematically empowered to deal with ADM systems.

While the report analyzes in-the-field examples from the EU’s member states, it also examines the way individual countries are addressing the social and political consequences. This reveals great differences between the member states with regard to investments, efforts to develop competences, and the existing depth of regulation. These gaps must be closed within the EU. Europe’s nations should find ways to share their experiences in dealing with ADM systems; they should learn from one another, and ultimately act in a united manner.

Closing the gaps between EU member states

In December 2018, Germany adopted a national strategy on artificial intelligence, following France, among others. However, both of these strategies lack concrete steps toward cross-border cooperation and the coordination of standards. While the documents show that there is a fundamental will to work together bilaterally, even this would fall short of what the situation demands.

Common European efforts are needed in order to utilize the new technologies in a way that promotes the common good. In this regard, not every automated decision necessarily rises to the same level of importance. Policymakers should focus primarily on ADM systems that touch on areas of citizens’ lives that are relevant to social inclusion. The algorithmic preselection of job candidates necessitates greater attention than do automatic spelling and grammar checks, for example. The goal must be a Europe that speaks with one voice to international actors when responses to the challenges and opportunities of automated decision-making systems are required.

At the invitation of MEPs Liisa Jaakonsaar (S&D), (Greens/EFA) and Michał Boni (EPP), we and AlgorithmWatch together presented our report to the European Parliament on 29 January. The video offers a summary of the discussion and a presentation of the results: