Council for Transparency and IAU present study on the use of algorithms in public bodies

To characterize the use and purposes that are given in the management of the State to algorithmic systems, and how the decision-making derived from their use affects citizens, the School of Government of the Adolfo Ibáñez University (UAI) – through the public innovation laboratory GobLab – and the Council for Transparency (CPLT) developed the study “Algorithmic Transparency in the Public Sector”.
From the institutions involved in this research commented that the increase in the use of this type of technology at the state level, for example in a process of admission to an educational establishment or to qualify for a social benefit, has led to a series of challenges. In this context, the so-called algorithmic transparency emerges as a relevant requirement to address in the effort to guarantee the protection of people’s rights, redefined from technological advances. 
The main results were announced within the framework of an activity organized by the entities and which included the presentation of the data, by Romina Garrido, lawyer, associate researcher of the GobLab and specialist in data protection; and commented by Danielle Zaror, Director optlA; Gloria de La Fuente, president of the CPLT, Senator Luz Ebensperger and Alberto Precth, executive director of Chile Transparente.
“The handling of large amounts of data, the minimum of responsibility and new possible forms of discrimination require not only transparency of acts and documents, but also of processes that support public decision-making,” said Romina Garrido. 
Likewise, Gloria de la Fuente also referred to the use of these systems in the management of the State “in the case of public bodies it can generate very relevant consequences for people’s lives, for example, access to a social benefit. For this reason, the Council signed this agreement that has allowed us to advance in an area where the scarcity of information predominates and is a line of work in which we must persevere due to the levels of ignorance that exist in this matter, for example, among officials, “said the head of the Council for Transparency, Gloria de la Fuente.  
Cadastre and transparency of algorithms
The research included the realization of a cadastre of the algorithmic systems whose use was reported by public bodies, an evaluation of the transparency of a subset of them and the elaboration of a normative recommendation for the State.
Within the framework of this process, 870 services were consulted, registering the participation of 33% of the required agencies. These reported a total of 219 systems, 59.4% of which were finally excluded, because they were not within the definition of automated decision systems object of the research.
This study allowed, among other things, to identify some difficulties associated with the conceptual approach of the systems that were sought to investigate, the identification of clear responsibilities in cases of local use of systems with centralized management, and also to know details of the publicly available institutional information of the declared systems. Among the main findings, he highlighted, with respect to the last point, that 80% of the systems did not have a history published on the websites of the public institutions that had implemented them.
As for the sectors in which these systems have been implemented, it was established that 38% of those declared are health, followed by economic affairs, with 13%, and public order and security, with 12%.
“This is empirical research that will finally allow other researchers, academics, civil society and citizens, to start thinking about the best way in which these systems can contribute to better public management and, by the way, address their risks, challenges and know their real impact.  It also shows an evident lack of transparency in communicating its existence, with findings that allow us to indicate that the capacities to be more transparent in the State exist and only need to be well directed,” explained Romina Garrido.
The data obtained with the study allowed to explore a set of 12 systems, to know their state of the art, starting with the basic information regarding their existence, identification, purpose, the data they use and their publicly available information.
Not only was inquired about the sArtificial intelligence systems, but also about those that use sequential algorithms that solve problems with public impact. These were analyzed under a selected algorithmic transparency standard. The results of the sample demonstrated a fortuitous advance of this matter driven by laws of transparency and participation of citizens in public management.
Also, that, despite the existence of information in few cases, this information is scattered and without a defined recipient. Among the systems chosen is a system of neural networks to detect fraud in medical licenses, a virtual assistant of FOSIS that has assisted in 2020 more than 270,000 people for the application of benefits delivered by that entity and Property Plan of the Agricultural and Livestock Service that from the monitoring of environmental variables advises on relevant aspects of the management of people’s properties.

Original source in Spanish

Related Posts

Add Comment