Artificial intelligence to improve interoperability in the European public sector

The European Union has placed the digital transformation of the public sector at the heart of its policy agenda. Through various initiatives under the Digital Decade policy programme, the EU aims to boost the efficiency of public services and provide a better experience for citizens. A goal for which the exchange of data and information in an agile manner between institutions and countries is essential.

This is where interoperability and the search for new solutions to promote it becomes important. Emerging technologies such as artificial intelligence (AI) offer great opportunities in this field, thanks to their ability to analyse and process huge amounts of data.

A report to analyse the state of play

Against this background, the European Commission has published an extensive and comprehensive report entitled "Artificial Intelligence for Interoperability in the European Public Sector", which provides an analysis of how AI is already improving interoperability in the European public sector. The report is divided into three parts:

1. A literature and policy review on the synergies between IA and interoperability. It highlights the legislative work carried out by the EU. It highlights the Interoperable Europe Act which seeks to establish a governance structure and to foster an ecosystem of reusable and interoperable solutions for public administration. Mention is also made of the Artificial Intelligence Act, designed to ensure that AI systems used in the EU are safe, transparent, traceable, non-discriminatory and environmentally friendly.
2. The report continues with a quantitative analysis of 189 use cases. These cases were selected on the basis of the inventory carried out in the report "AI Watch. European overview of the use of Artificial Intelligence by the public sector" which includes 686 examples, recently updated to 720.
3. A qualitative study that elaborates on some of the above cases. Specifically, seven use cases have been characterised (two of them Spanish), with an exploratory objective. In other words, it seeks to extract knowledge about the challenges of interoperability and how AI-based solutions can help.

Conclusions of the study

AI is becoming an essential tool for structuring, preserving, standardising and processing public administration data, improving interoperability within and outside public administration. This is a task that many organisations are already doing.

Of all the AI use cases in the public sector analysed in the study, 26% were related to interoperability. These tools are used to improve interoperability by operating at different levels: technical, semantic, legal and organisational. The same AI system can operate at different layers.

• The semantic layer of interoperability is the most relevant (91% of cases). The use of ontologies and taxonomies to create a common language, combined with AI, can help establish semantic interoperability between different systems. One example is the EPISA60 project, which is based on natural language processing, using entity recognition and machine learning to explore digital documents.
• In second place is the organisational layer, with 35% of cases. It highlights the use of AI for policy harmonisation, governance models and mutual data recognition, among others. In this regard, the Austrian Ministry of Justice launched the JustizOnline project which integrates various systems and processes related to the delivery of justice.
• The 33% of the cases focused on the legal layer. In this case, the aim is to ensure that the exchange of data takes place in compliance with legal requirements on data protection and privacy. The European Commission is preparing a study to explore how AI can be used to verify the transposition of EU legislation by Member States. For this purpose, different articles of the laws are compared with the help of an AI.
• Lastly, there is the technical layer, with 21% of cases. In this field, AI can help the exchange of data in a seamless and secure way. One example is the work carried out at the Belgian research centre VITO, based on AI data encoding/decoding and transport techniques.

Specifically, the three most common actions that AI-based systems take to drive data interoperability are: detecting information (42%), structuring it (22%) and classifying it (16%). The following table, extracted from the report, shows all the detailed activities:

Download here the accessible version of the table

The report also analyses the use of AI in specific areas.  Its use in "general public services" stands out (41%), followed by "public order and security" (17%) and "economic affairs" (16%). In terms of benefits, administrative simplification stands out (59%), followed by the evaluation of effectiveness and efficiency (35%) and the preservation of information (27%).

AI use cases in Spain

The third part of the report looks in detail at concrete use cases of AI-based solutions that have helped to improve public sector interoperability. Of the seven solutions characterised, two are from Spain:

• Energy vulnerability - automated assessment of the fuel poverty report. When energy service providers detect non-payments, they must consult with the municipality to determine whether the user is in a situation of social vulnerability before cutting off the service, in which case supplies cannot be cut off. Municipalities receive monthly listings from companies in different formats and have to go through a costly manual bureaucratic process to validate whether a citizen is at social or economic risk. To solve this challenge, the Administració Oberta de Catalunya (AOC) has developed a tool that automates the data verification process, improving interoperability between companies, municipalities and other administrations.
• Automated transcripts to speed up court proceedings.  In the Basque Country, trial transcripts by the administration are made by manually reviewing the videos of all sessions. Therefore, it is not possible to easily search for words, phrases, etc. This solution converts voice data into text automatically, which allows you to search and save time.

Recommendations

The report concludes with a series of recommendations on what public administrations should do:

1. Raise internal awareness of the possibilities of AI to improve interoperability. Through experimentation, they will be able to discover the benefits and potential of this technology.
2. Approach the adoption of an AI solution as a complex project with not only technical, but also organisational, legal, ethical, etc. implications.
3. Create optimal conditions for effective collaboration between public agencies. This requires a common understanding of the challenges to be addressed in order to facilitate data exchange and the integration of different systems and services.
4. Promote the use of uniform and standardised ontologies and taxonomies to create a common language and shared understanding of data to help establish semantic interoperability between systems.
5. Evaluate current legislation, both in the early stages of experimentation and during the implementation of an AI solution, on a regular basis. Collaboration with external actors to assess the adequacy of the legal framework should also be considered. In this regard, the report also includes recommendations for the next EU policy updates.
6. Support the upgrading of the skills of AI and interoperability specialists within the public administration. Critical tasks of monitoring AI systems are to be kept within the organisation.

Interoperability is one of the key drivers of digital government, as it enables the seamless exchange of data and processes, fostering effective collaboration. AI can help automate tasks and processes, reduce costs and improve efficiency. It is therefore advisable to encourage their adoption by public bodies at all levels.