Automating Welfare

Across Europe, the welfare sector is facing increasing demands and shrinking resources. Calls to mitigate social hardships more effectively and increase public administration`s efficiency suggest introducing automated decision-making. There is a vision that welfare can be radically transformed through data-driven decision-support and enhancement with artificial intelligence. In many areas, automation is thought to increase productivity with less effort and lower costs. What happens to human flourishing, social justice, and equity when this logic is deployed to support decisions in the welfare sector? AUTO-WELF investigates the extensive implementation of automated decision-making (ADM) in the welfare sector across Europe. It is the first project to provide an analysis of automated welfare provision across different European welfare regimes. It examines the implications of algorithms and artificial intelligence for the future of European citizens and societies. Emerging ADM systems perform many functions in societysome of which are mundane and perceived as relatively unproblematic and beneficial for citizens. Still, others can disrupt and change living circumstances. Data-based infrastructures for public administration have an impact on human agency within complex socio-technical systems. This prompts many ethical and political questions regarding transparency and accountability, as well as systemic biases and social inequalities. ADM systems are shaping not only welfare provision and public management but also how citizens encounter and are addressed by the state and how the relationship between citizens and the state is transformed. AUTO-WELF foregrounds the perspective of people implicated in the automation process the system engineers and designers, the caseworkers who collaborate with ADM systems in making decisions about welfare and service provision, and the people whose data feed the systems and who are targeted in processes of enhancing welfare through automation. The project will develop groundbreaking knowledge on the consequences of automating welfare in two domains: a) core welfare services, such as employment services, healthcare, and social benefits provision, and b) automation of communal welfare infrastructures and services, including smart city and smart village initiatives that aim at providing automated social infrastructures for community building and urban development. We explore these domains across eight European countries: Austria and Germany, Italy and Portugal, Estonia and Poland, and, Denmark and Sweden represent different types of the welfare state and different stages of automated decision-making. With its interdisciplinary approach, AUTO-WELF will provide an in-depth and cutting-edge understanding of automating welfare from a European perspective. The project will deliver highly relevant insights into how automated decision-making can support but also harm human flourishing.

With the use of large datasets and increasingly powerful computational models, we are observing a new phase in the digital transformation of the welfare state, driven by artificial intelligence (AI), algorithmic decision support and the growing use of data analytics. AUTO-WELF analysed the implications of automated systems in eight European countries: Austria, Denmark, Estonia, Germany, Italy, Poland, Portugal, and Sweden. Research in Austria and Portugal was carried out by a team led by Dr. Doris Allhutter based at the Austrian Academy of Sciences in cooperation with a European consortium directed by Prof. Anne Kaun. Currently, a multitude of data-driven systems are being developed by in-house departments and in public-private partnerships. The delegation of decision-making to algorithmic systems and AI-often accompanied by other organisational changes-transforms the way citizens encounter the state and how power is exercised. Systems are implemented with the goal of creating cost-effective and objective, data-driven public policy. However, existing research points to data bias, flawed models and a lack of transparency. Opaque shifts in how decisions come together directly affect people's livelihoods, and the scope of their rights and obligations. In Austria and Portugal, we observe the gradual testing and dissemination of data-driven systems in various domains. First, there is a proliferation of algorithms in public benefit provision: both countries, for instance, automate welfare fraud detection in public social security. Second, essential services are increasingly 'datafied': Lisbon has an urban management and intelligence centre, which runs a data integration platform used by public service providers, e.g. emergency and transport services. Third, interconnected data infrastructures collect data from federal and municipal levels for various public uses: in Austria, an open government data platform collects metadata from the decentralised data catalogues of public administration, the Austrian Micro Data Centre provides administrative data to accredited research institutions, and cities build digital GeoTwins for urban management. These and similar examples across the eight countries occur like an ongoing beta testing that leads to growing data care duties. Information on how data-driven systems are used in social policy is however fragmented, partial, and difficult to access. Even though it is challenging to systematically document their social effects, AUTO-WELF shows that the implementation of such systems is often accompanied by a hidden redefinition of welfare state goals. Data-driven systems fundamentally reshape knowledge and work practices in public institutions and data is increasingly utilised as a strategic, infrastructural resource. We have observed trends characteristic of neoliberalism such as the individualisation of social issues; the curtailing of legally guaranteed social rights through decisions hidden in the technology; the prioritisation of welfare surveillance in place of universal, high-quality social services; and the marginalisation of citizens' and civil servants' voices in designing reforms and shaping sociotechnical processes.