Artificial intelligence (AI) should never replace democratic decision-making, but it can be used to help legislators and staff make legislation clearer, better designed to achieve the desired outcomes and easier to implement. It is an important current topic not only globally but also at the European Parliament.
Quality over speed
The conference on 'AI for legislative work' at the European Parliament was organised by the Panel for the Future of Science and Technology (STOA) and took place on 4 November with a number of experts.
The Parliament continues to seek ways AI can be used to provide quality support to legislative work. However, as both European Parliament President Metsola and Secretary-General Chiocchettihighlighted in their addresses, responsibility for the use of AI-generated outputs remains with elected politicians. 'AI cannot ultimately replace the judgment and critical thinking needed to make tough decisions', underlined President Metsola.
AI for different aspects of policymaking
At the event, four examples of applications of AI, covering various parts of the policymaking cycle, from different EU Member States were presented, along with their pros and cons and the governance and other practical issues they raise.
Monica Palmirani gave an overview of the system used in the Italian Chamber of Deputies to structure and check the consistency of proposals and amendments with national and EU law, as well as jurisprudence.
Emmanuel Cortes explained the system used by the French Prime Minister's office to summarise and group amendments in the French Assembly.
Veera Heinonen presented a Finnish application designed to help with understanding EU laws in force and simplify compliance for public authorities and businesses.
Finally, Thomas Hemmelgarn showed participants the Joint Research Centre's exploration of AI personas to represent real data on people's views as an initial sounding board to help gauge the impact of proposals on different segments of society.
Data, control and 'explainability'
In the subsequent panel discussion, there was agreement that having high-quality and secure data is the cornerstone of any successful AI application. The need for control over the models used, whatever their origins, as well as the ability to provide verifiable sources to back up results were also underlined.
The event was extremely well attended, highlighting the interest, inside and outside the Parliament, in the use of AI in policymaking.
