Artificial Intelligence in the Latin American Public Sector
Across the world, governments are incorporating artificial intelligence (AI) into their arsenal of tools to improve efficiency, make better policy decisions, and increase engagement with the public. This trend holds true for many nations in Latin American and the Caribbean (LAC). A new report by the Organisation for Economic Co-operation and Development (OECD), an international policy analysis think tank, explores how LAC governments are incorporating AI into their processes for a more responsive public sector. We (Edgelands Institute) spoke with Ricardo Zapata, an analyst at the OECD and one of the co-authors of the report, “ The Strategic and Responsible Use of Artificial Intelligence in the Public Sector of Latin America and the Caribbean “ to learn more about how LAC governments are using AI and what this means for the future of responsible governance, public security, and digital rights in the region.
What are some of the most common ways that a government may use AI?
Governments will generally approach AI to improve efficiency. Concretely, this will mean automating simple tasks or upgrading the speed and quality of public services. Most of the visible use cases belong to this category. In some other cases, governments will also look to improving the policy design process by getting in-depth knowledge out of large quantities of data (and most governments are already good at producing a lot of data) and making more informed decisions. This is an area where AI can deliver good results, but that also calls for stronger ethical approaches in the whole cycle, from the collection of data to its processing. Finally, AI is being used also to enhance communication and engagement with the public, generally through chatbots, matching tools (producing better recommendations according to citizens’ needs or characteristics), or improving the abilities to better understand opinions and perspectives of citizens at scales that were previously unfeasible. Good to note that similar uses happen the other way around when citizens use AI to better understand what their governments are doing.
The report states that Colombia has emerged as a leader in the LAC region for its use of AI. Why is that?
Colombia is building strong AI policy and producing very interesting use cases. On the policy side, we documented various instruments and levers that secure strong governance of AI in the public sector. In the first place, its AI strategy is the only one in the region to have the whole set of enablers that can help drive implementation. With enablers we mean having objectives and specific actions, measurable goals, responsible actors, time frames, funding mechanisms, and a monitoring instrument. Although they do not guarantee successful implementation, having them in place can improve its performance, especially when we consider Colombia is a large country and the strategy has been made to endure across administrations.
Second, we analysed 15 policy levers that can improve the implementation of AI projects in the public sector and, in most of them, Colombia stood out as a regional leader. For example, when looking at the efforts to develop a responsible, trustworthy and human-centric AI, Colombia’s ethical framework was well aligned with our reference instrument, the OECD AI Principles. When looking at funding, Colombia’s national strategy was unique in the region in having an explicit funding mechanism. When looking at data governance or spaces for experimentation, it was also among those countries with the most complete instruments and practices.
This whole evaluation exercise did not intend to create a ranking, but to identify strengths among countries to enable others to learn from their practices and lessons. Our study does not say who the best is and neither who has the greatest impact. At this stage, we can compare what countries are doing with the policy frameworks we are developing and, being AI such a recent trend, it is possible that these frameworks evolve in the coming years. So, being a regional leader, in this case, means that countries should take a special look to what Colombia is doing.
Could you give an example of how Colombia uses AI in the public sector?
Aside from all the policy developments I described, Colombia is also making some interesting implementations in the public sector. I think PretorIA’s case stands among the most interesting we documented. This is a project developed by the Constitutional Court to help in the selection of key tutelas (i.e. Constitutional Action for the protection of fundamental rights) to set legal precedents on the provision of fundamental rights. The Court receives more than 2000 tutelas each day, so a solution to make this process more efficient can add a lot of value. PretorIA automatically reads and analyses all plaints, detects and predicts the presence of predefined criteria, and intuitively presents reports and statistics. It serves as a tool for judges, so it ensures there is a human in charge of the decision-making process.
What I found the most interesting about this case is how it was developed. Initially, it was an adaptation of Prometea, an AI system developed in Argentina to help justice providers. It acted as a virtual assistant that predicted case solutions (based on previous cases and solutions) and helped provide information required to assemble case files. When the Constitutional Court first announced the project in 2019, it sparked citizen participation and dialogues with many stakeholders due to concerns about the opacity and black-box effect of this system, especially when dealing with a process made to guarantee fundamental rights. Citizen participation then led the Court to transform the whole project into what we now know today as PretorIA, a system that uses topic modelling technology instead of neural networks, making it more explainable, interpretable and traceable. The institutional reaction to external voices is a nice example of responsible actors being held accountable, making the necessary changes to assure the trustworthiness of the AI system.
What is a “trustworthy, human-centered” approach to AI and how does it differ from traditional uses? How does Colombia rank in its use of human-centered AI?
Our reference framework to assess trustworthy and human-centered approaches are the OECD AI Principles, which intend to promote the development of innovative AI that respects human rights and democratic values. Coming from an institution like the OECD dedicated to sharing best practices, establishing international standards, and improving public policies in democratic and market-oriented economies, the report seeks then to interrogate how public sector AI systems are integrating these values. I think it is hard to say what a traditional approach is. What I would stress is that technology is not value-neutral, so every development embeds a vision of how the world should look like. Many systems we know today have put their focus in efficiency, profit, or control. As citizens and institutions, we should therefore permanently ask how we want technology to be and what values do we want it to reflect.
In this regard, Colombia has done a great effort issuing the AI Ethical Framework, aligning with the OECD AI Principles. Our report goes more in depth and looks at particular aspects of the responsible, trustworthy and human-centric approach, so we look at fairness and bias mitigation, transparency and explainability, safety and security, and accountability. In all of those, Colombia’s AI Ethical Framework gives the country good capacities to promote this type of approach. Where we see room for improvement is in setting guidance and methods for understanding user needs and for building diverse AI teams. Of course, all this concerns the existence of policy frameworks, which is in itself a great progress, so their application and impact will be a matter of future research.
How could human-centered AI help people coexist better in cities? Could you give an example of how this technology could be used in a city planning or public security intervention?
Cities are a context where digital government projects can have a more tangible and visible impact for citizens. Implementing a trustworthy and human-centric AI in domains like security, where we are dealing with the privacy and integrity of persons, could allow society to be confident on the right applications of the technology and held responsible actors accountable for its results. It could also help us decide where we do not want to apply it, which is a completely valid choice. We talk a lot about the benefits of AI, but sometimes a risk-based approach can say: these domains should remain tech-free. For instance, facial recognition, which has been banned in some places and remains a debated topic. It could also help us discover that AI is not the best solution, as it happened with Predpol in Montevideo.
At the urban level, many authorities in Latin America are using AI systems to analyse the enormous amounts of real-time and historical mobility data to improve traffic management, detect risks, or model future scenarios. Some are also using surveillance cameras and devices together with AI algorithms to track criminal activity, identify persons, and analyse historical data. Our report highlights that there are still two main challenges in this last area: establishing the necessary safeguards when processing sensitive personal data (e.g. biometric data), and defining clear frameworks for the use of these technologies to prevent possible abuses such as the profiling and persecution of political opponents or protesters.
We certainly need more active discussion to shape urban surveillance technologies. There is still a lot of progress to be made. In the end, that is one of the points of incorporating democratic values in the development of AI, that we can collectively shape these technologies and their applications.
Originally published at https://edgelands.institute.
