How does AI in public decision-making affect the formation of public values? A study of professionalism and accountability in labor market policies

Research problem and specific questionsAI is on the verge of reforming how policy is implemented in public administration. The rapid development of AI technology raises several important questions.

While probable to strengthen efficiency and equality in decisions and policy output, AI is likely to affect other important public values as well. However, empirical knowledge on this is very scarce.

The increasing use of AI in public decision-making allows algorithms to intercept with the discretion of public officials.

This is likely to affect two values closely linked to discretionary practices in public administration: professionalism and accountability. Our proposed project examines how these values are affected by the use of AI in public decision-making. For example, how is human reasoning, and professional judgement and ethics, reshaped by the adoption of AI systems?

Who can be held accountable if decisions are made by a machine?Data and methodWe examine an AI decision-making tool implemented by the Swedish PES by conducting three empirical studies.

In study 1, we conduct a document study and interview central officials at the ministry of labor and the PES to examine how AI is governed in practice. Study 2 examines how AI affects the professional practices of public officials. Through in-depth interviews with caseworkers, we analyze how they reason and act in relation to the tool.

For instance, how the tool affects the use of professional judgement, and how the caseworkers view their own responsibility for decisions informed by the tool. Study 3 is a survey experiment conducted with PES caseworkers.

The aim is to deepen our understanding of how professional judgement and the output from the decision-making tool are balanced. This approach contributes with knowledge on what role algorithmic recommendations play in actual decisions.

Societal relevance and utilisationThe results of the project are expected to be of great importance for policymakers, professionals, and citizens who face AI in public administration in different ways.

Our focus on the experiences of caseworkers will provide systematic and in-depth knowledge from a perspective that is largely missing today.

Moreover, our findings can serve to inform policy design and governing efforts.Plan for project realisation The project runs for three years and two researchers take part. In this time, we conduct three empirical studies using different methodologies and develop a theoretical framework.