The Social Sector Manager of the Inter-American Development Bank, Marcelo Cabrol, talks about fAIr LAC, a pioneering project in which governments and companies will work on applying AI to the well-being of society
Marcelo Cabrol has spent nearly six months traveling halfway around the world in search of answers. His boss, the president of the Inter-American Development Bank (IDB), Luis Alberto Moreno, has tasked him with exploring what the future might look like for an institution devoted to financing projects in Latin America. The big question this economist of Argentine origin has set out to answer is: can technology be fair and fulfill its promise of improving social policies? To find out, the IDB will launch fAIr LAC in October, a pioneering project in which governments and technology companies will work together in search of the formula that will allow artificial intelligence (AI) to contribute to social well-being.
Cabrol, who previously headed the bank’s communications, speaks passionately about technology and insists strongly on the importance of ethics. “In education, health, gender and diversity policies… very personal data is handled. That is why we have to talk about ethics and regulation when we talk about AI”, he explains. “At the end of the day, technology can end up breaking the social contract”.
Which companies have already said yes to participating in fAIr LAC?
Google, Microsoft, NTT, and Telefónica, which, for now, is our only European partner. We are going to bring together public and private data.
Are there precedents for anything like this in the field of social policy?
No, no one has done it until now, and that is why it has taken us so long to launch this program.
Do governments agree to share their data with these tech companies?
Some do, yes. We have already closed programs with Uruguay and Mexico. In the State of Jalisco there is a project under way for the early detection with AI of diabetic retinopathy, the leading cause of blindness in the country.
What does a company like Google gain from all this?
It is no secret that it has not been a good year, and I think they could gain legitimacy. They can also access knowledge about markets and populations they currently do not reach. But to achieve the promises of artificial intelligence, the ethical side of all this must be discussed.
But if citizens are not very informed about the data these companies hold…
If an AI-based model is applied to public programs, citizens start demanding accountability. In Argentina, for example, a study was carried out that predicted the probability of a teenage pregnancy with 75% accuracy. When society started asking questions, it turned out the model was based only on responses from girls from poor families. It was very biased and could not explain how it reached its conclusions.
Doesn’t that happen with all algorithms?
Yes, but we are talking about teenage pregnancies, not selling a pair of pants. If you can establish with certainty that a girl is going to become pregnant and you have no policies to prevent it, then it is the government that has an ethical problem. The same goes for a model capable of predicting which students will drop out of school. This is where the change comes in: how to incorporate policy solutions to all these problems.
Are these evidence-based policies?
They are much more than that, because you are acting on the outcome before it happens. That is what is interesting. For the IDB it is essential to help countries complete this work and make investment in health and education more equitable.
How does the IDB play that role?
By creating evidence. We need a lot of experimentation. And that means ensuring there is no bias, that there is transparency, data privacy, and making sure the program works. Our vocation is not pilot programs but taking these solutions to scale.
Even with the diversity of Latin America?
We believe so. It is much cheaper to adjust the model than to start from scratch.
An example of a project the bank is thinking about?
We want to implement image-based breast cancer diagnosis using AI in Amazonian communities that lack access to this type of service. The idea is not new, but we are trying to think it through ethically. To begin with, who stores the information, and where?
And the application of AI to education in Latin America?
I have just been in Finland. We know the keys to their success: they have very good teachers, personalized learning, and 8,000 euros of spending per child. We cannot wait to have all that. I wish we could, but I do not think I will live to see it. Technology makes it possible to skip a few steps.
Personalizing education…
That is the great promise we are betting on. YouTube is already making decisions for you when it suggests the next video to watch. The idea is to use that same approach. If each child learns differently and has different abilities, with AI we should be able to design more tailored content.
This is closely related to the skills needed later to look for a job.
Yes, but there are also the things you do not see. When you fill out a job application form, the questions are not designed to tell you: “You would be very good at this job”. But if I complement that information with socio-emotional skills, I will have a clearer picture of what your next profession would be and how you should train for it.
Is leaving so many decisions in the hands of AI complicated?
That is the great challenge. Today the big criticism we level at technology is: Who decides the model? Do we know if they are manipulating the results? When we talk about AI we go on alert: “Careful, here comes a technology that is going to make decisions with a moral framework that nobody knows and that is not explained anywhere”. It is an ethical and moral problem too. That is why we are thinking of an interdisciplinary approach: we need political scientists, anthropologists, sociologists, philosophers… That is not exactly the profile of the bank’s staff.
Do you plan to integrate these professional profiles into the bank?
Yes, those are the profiles we are looking for. It is not that the IDB wants to become the champion of ethics. Our intention is to work with governments and tech companies. It is true that all the big tech companies have their ethical principles. However, we believe it is important to maintain an industry-wide conversation, beyond what legislation imposes. It is an ambitious project, but a necessary one.
What is the budget for the fAIr LAC project?
We have 15 million, expandable by five more, and we want to launch 20 pilot programs.
And how is this policy approach fitting in within the IDB?
It will not necessarily mean a radical change, although the future of the bank will not lie in financing. Today, the IDB lends 13 billion dollars and does it very well. It could continue like this, giving loans for hospitals, schools… and it would still have a business. However, much of the investment we make has to do with training people in these sectors. What are the non-traditional ways of doing it? fAIr LAC will tell us. Will this happen tomorrow? No, but I have to get ready.
