The AOC creates a transparency register of artificial intelligence algorithms at the forefront of the State
Algorithmic transparency is a topic of increasing importance in our society, since the artificial intelligence (AI) algorithms they are increasingly used to make decisions that affect people's lives.
To ensure algorithmic transparency, it is essential that AI algorithms can be understood by the people who use and affect them, so that they can understand how a certain decision was reached and why that decision is made; as well as that the algorithms can be reviewed to detect errors or unintentional biases and improve their accuracy.
Transparency is therefore an essential requirement for achieving and maintaining user trust in AI systems, and a crucial issue for promoting responsible, safe and societally fair development and use of this technology.
To comply with the requirement of transparency of AI systems used in public administrations the existing regulations do not establish, today, any "concrete" legal mechanism or instrument; although the AI Regulation that is being prepared by the Council of the EU points out, as a first recommended action, to "register and document" the AI systems used by the organizations that may be subject to this regulation. The rest of the regulations speak of encouraging the implementation of "mechanisms" so that the algorithms involved in decision-making take into account transparency criteria, without specifying any; or to publish on the website the list of automated administrative actions (AAA) with information on the rules applied and how they work.
What are we doing to ensure transparency of AI algorithms in the AOC?
As a means of achieving transparency in the use of AI algorithms, the AOC proposes the publication on the Transparency Portal of explanatory sheets for each algorithm used.
This is a leading initiative of algorithmic transparency in the public sector of Catalonia and the State.
This initiative aims to be a tool for:
Provide users with an overview of the algorithmic systems used in the AOC, including information on the characteristics and results of their operation, the risks and biases that have been identified and the measures that have been applied to correct them , and the body before which particular decisions can be appealed or its general operation challenged.
Make it easy for everyone to give their opinion and participate in the creation of people-centered algorithms.
Thus, for each AI algorithm employed in the AOC's digital administration services, we will make a fact sheet with general information about the purpose and operation of the algorithm, and then give more detailed technical information. There will also be the ability to submit feedback to help the development team make the algorithms we use better, fairer, and more accountable.
Below you can consult the first file of the initiative published on our Transparency Portal. This is the tab for conversational chatbots that we apply to the VALid, idCat Certificat services and, soon, to the Transparency Portal itself.
Want to learn the secrets behind the AI algorithms of AOC's conversational chatbots?
The transparency registry of artificial intelligence algorithms has been mainly inspired by the excellent practical experiences of city registries Amsterdam and from Helsinki who were leaders in Europe.