[Return to ELISE's Artificial Intelligence in the public sector page]

This report led by the Joint Research Centre as part of the AI Watch project, has received the contribution from the ELISE action in the mapping and analysis of the use of AI in EU public services. Know more on AI in the Public sector in the AI Watch dedicated page.

 

Public administrations across the European Union are increasingly exploring the application of Artificial Intelligence to improve public services, policy making and internal management. This report fills an important gap in the current research on the implementation of AI technologies in government and public services, as this field has not gained the same attention as AI for the private sector, despite the potential for improving citizen’s well-being. Instead, governments are often solely seen as the regulator of AI or as a facilitator for AI development in the private sector. However, the government is also very suitable as a user for AI to improve their operations.

Fuelled with the power of data, the diverse set of AI technologies are expected to play a positive role in making governments work more effective and efficient, transforming how public services are delivered and how policy is designed. Although, the positive impact of AI is far from straightforward and should not be taken for granted, due to significant adoption barriers as well as the need to avoid potential negative effects of AI deployment. Consequently, there is a strong need to assess what the positive and negative consequences of AI use in public services are. This report on AI in public services lays the groundwork for future understanding how AI is being deployed in public services and what its consequent effects are. Thus, it proposes an approach which takes into account an ex-ante and ex-post situation for impact assessment.

The report includes an exploration of the current literature on AI in public services, an analysis of 230 identified AI use cases within the EU27 + UK, Switzerland and Norway, illustrative case studies, the results of a survey among Member States, an analysis of the national AI strategies regarding policy initiatives for AI use in the public s

ector and the proposed approach to assess the impact of AI in public services.

The first landscaping exercise resulted in a first inventory of 230 illustrative AI use cases, highlighting the variety and interest of government organisations to experiment with AI. Common AI typologies found in the current inventory include the use of Chatbots or other Digital Assistants to provide information or services to citizens or AI providing data-based predictions, through the recognition of patterns in datasets. The inventory should act as a suitable starting point for more in-depth studies on high impact use cases and comparative studies.

Despite the transformative expectations of AI, more than half of the AI solutions in the inventory suggest merely incremental or technical changes to the administration adopting them. More disruptive or even radical changes in the way organizations function are far less common and often under scrutiny or criticism, suggesting that the research on AI in government require a long-term perspective. 

This report is merely the starting point of a new research field, as many open issues, questions and paradoxes remain. The term AI remains poorly defined, with many different types of technologies and applications being regarded as AI even when they function very differently in practice. Furthermore, it remains unclear how the existing challenges in adoption can be overcome, showing the need to learn from successful examples and to gather best practices. Future research should go more in-depth to discover the conditions for successful AI implementation as well as the positive and unintended consequences of their use.