For decades, the answer to the question of who decided in the State was limited to an employee, a leader or a minister, but today the algorithm can increasingly be included.
Algorithmic and Artificial Intelligence systems now help to define priorities in inspections, detect fraud, organize waiting lists, assign support or flag risks. They are useful, necessary tools that influence people’s lives. And everything that influences rights, opportunities or duties must be subject to public light.
Hence the importance of a Public Registry of Algorithmic Systems and AI in the Public Sector. It is not a technological whim, nor a bureaucratic excess, it is democratic transparency applied to the 21st century.
Its purpose is to allow any citizen to know which systems are being used by the State, what they are for, on what legal basis they operate, what type of data they use and what effects they produce. If an algorithm contributes to deciding who is supervised or who receives support, this cannot be an administrative secret.
However, transparency does not mean irresponsible exposure, as a well-designed record distinguishes what should be public from what should be reserved. The citizen must know the purpose of the system, the responsible entity, the degree of automation, the existence of human intervention, the data categories used and the dispute mechanisms. Sensitive technical details, vulnerabilities or legitimate commercial secrets must remain accessible only to control and audit bodies.
It is equally important to explain that not all algorithms are the same. There are systems that only support human decisions, others classify or prioritize processes and others can produce automated decisions. The record must make clear the degree of human intervention involved, as a citizen has the right to know whether they are facing a technical recommendation or an automated decision.
Another essential point is risk assessment. A system that organizes internal schedules does not have the same impact as one that influences access to social benefits. The registry must indicate the assigned risk level and summarize impact assessments, including those on data protection, and make public the main mitigation measures.
The issue of data also requires clarity. This is not about publishing databases, but about indicating which categories are used (such as tax, identification or health data, among others), what is the legal basis for the processing and what are the retention periods. This strengthens trust and allows for informed scrutiny.
Furthermore, systems are not static. They evolve, are updated and may reveal flaws or biases. The log must include information about versions, audits performed, aggregated performance metrics and relevant incidents. If a significant error or important correction has occurred, the public should know about it.
Creating a record like this is not a technical mystery. It requires a central database, a searchable portal, submission mechanisms by public entities and legal and technical validation before publication. Requires integration with public procurement records and personal data processing inventories. And it requires clear governance rules, to know who registers, who validates and who supervises.
But the essential thing is not the technology, it is the democratic culture that gives it meaning. In a Rule of Law, power, even when exercised by code, must be visible, explainable and contestable. Therefore, transparency today must also cover the systems that structure decisions.
Public trust is born from clarity and transparency, so in the era of Artificial Intelligence, democracy begins by knowing which algorithms govern us and how they do it.
For all these reasons, today, at Campus XXI, in Lisbon, a workshop of reflection on The Transparency of Artificial Intelligence in the Public Sectororganized by APDSI – Association for the Promotion and Development of the Information Society, with support from ARTE – State Agency for Technological Reform, within the scope of the III Open Administration Plan.
Leave a Reply