Skip to main content

How AI helps the Italian Senate manage amendments

Innovation tracker | Issue 12 | 06 Oct 2022

Senate clerks use “Gestore Emendamenti” (GEM), an amendment management system. The table above shows similarity clusters for a group of amendments

In the Italian Senate, amendments are presented electronically, but not always for the purpose of changing proposed legislation. Often, members – particularly those from opposition parties – propose amendments as a way to advertise alternative policies or to challenge the government. In extreme cases, they may present large numbers of amendments, differing by just a few words, in order to slow down the legislative process – a practice known as “filibustering”. This can prove challenging for the Senate staff responsible for analysing and organizing amendments and scheduling voting.

How AI assists

The IT department has developed various document management and automation tools for Senate staff. A new AI-powered system for managing amendments was introduced as part of an experiment to handle workload peaks, such as those caused by filibustering. This system, which uses “text clustering algorithms”, speeds up the detection of groups of similarly worded amendments.

This experimental feature is already demonstrating its value. According to Carlo Marchetti, Head of the Senate’s Information Systems Development Office, the process is almost instant. But a human eye is still needed to review, approve, modify and integrate the system’s results. No decisions are taken without supervision. In this sense, the AI technology is there to assist human beings, not to replace them.

What's next

The Senate developed this feature in collaboration with the Institute of Legal Informatics and Judicial Systems, which is part of the National Research Council of Italy (CNR). More technical information about the clustering algorithms, and how they are integrated into the Senate systems, can be found here. As part of further research into new tools and algorithms to support core parliamentary activities, the Senate is exploring the possibility of detecting not just textual but also semantic similarities, and of identifying related bills that could be similarly affected by the amendments.