Automated decision-making systems are analyzed at four institutions to make their use more transparent: how they operate in the allocation of a bonus or in the socioeconomic classification for the Social Household Registry.
UAI and Chilean Transparency Council lead pilot at public agencies for first regulations on algorithmic transparency in Latin America
17 de octubre 2022
Automated decision-making systems are analyzed at four institutions to make their use more transparent: how they operate in the allocation of a bonus or in the socioeconomic classification for the Social Household Registry.
The initiative is part of a process that will culminate in the drafting of a General Instruction on Algorithmic Transparency by the Chilean Transparency Council, which will stipulate the information that public institutions must communicate about these systems and their use.
Seven automated decision systems with a high impact on the provision of services to citizens are part of the pilot currently being conducted in Chile by the Chilean Transparency Council (CPLT, Consejo para la Transparencia) and Universidad Adolfo Ibáñez (UAI) – through the Ethical, Responsible and Transparent Algorithms project supported by the IDB Lab – to gather information about algorithms and their use in the decision-making process of four public agencies.
Based on these and other inputs that will be collected through various participatory instances, among them workshops run by UAI, the first General Instruction on Algorithmic Transparency will be prepared by the CPLT.
This tool seeks to make the operation of algorithms in decision-making processes by the public sector transparent, for instance, the calculation of the value of a bonus quota, the socioeconomic classification for the Social Household Registry or the School Admissions System.
The entities participating in this pilot are the Health Superintendence, the Social Security Institute, the Social Evaluation Undersecretariat, and the Superintendence of Insolvency and Re-Entrepreneurship.
The CPLT and UAI highlighted the relevance of the current phase as input for the document’s construction, which includes automated systems currently used in decision-making processes carried out by these public institutions within the scope of their functions, several of them associated with extensively used services or public policies, and therefore, with great impact on citizens. Some of them are the system that allows calculating the quota value of the live-born child bonus of the Social Security Institute and the algorithm that calculates the socioeconomic classification of the Ministry of Social Development and Family.
The initiative contemplates making information available about the ways in which these systems operate in order to obtain an understanding, in plain language, of their functioning and other relevant aspects. In the words of the CPLT president, Francisco Leturia, these systems “decide on fundamental aspects of people’s lives; whether a subsidy is granted or not may depend, to a greater or lesser extent, on the result or input provided by an algorithm”. In this regard, he added: “The role of the Council for Transparency is to safeguard the right of any person who may need it to access information held by public institutions, which includes disclosing how decisions are made in the public sector and based on what information, even when they use automated or semiautomated decision-making systems that can many times be opaque, unknown or complex to a majority”.
María Paz Hermosilla, director of GobLab UAI, shared that a previous study (see www.algoritmospublicos.cl, spanish) observed that “80% of the algorithmic systems being used by the State do not offer information about their functioning, the origin of the data or other aspects. Therefore, the motivation behind this work with the Council is related to the State’s need to be accountable and respect people’s rights”.
What does the pilot entail?
The pilot’s participating institutions published information on the State Transparency Portal about a series of automated decision systems they use to provide services to citizens. The purpose is, on one hand, to generate lessons learned within the organizations and to detect potential difficulties they had in collecting the information and making it available and, on the other, for users to review the publications and evaluate their usefulness, pertinence and whether they are easily understandable or not through workshops to be developed by UAI and the CPLT.
This is a first step towards raising transparency standards in the use of these types of systems by the State, thus contributing to the legitimacy and reliance of the technological solutions. This is one of the phases contemplated by the Council in this process, through which it will draft the first regulation on this matter by means of the General Instruction on Algorithmic Transparency. This document, unprecedented in Latin America, seeks to ensure higher levels of transparency, for instance, regarding their actual existence and functioning. Another objective of the work is for people not specialized in these issues to have information about these technologies or tools available in plain language, bringing transparency closer to people.
A process open to citizens
The CPLT and UAI underscored the participatory nature of the process that will culminate in the General Instruction and also informed about some milestones open to the civil society and citizens. For instance, prior to this phase, roundtables were held with over 20 public institutions, and various civil society organizations will be invited to workshops in the coming days to present some developments to them. Additionally, they expect to submit the content proposal to an open public consultation by the end of 2022.
