EL PAÍS offers the América Futura section for free for its daily and global informative contribution on sustainable development. If you want to support our journalism, subscribe here.
When Facundo’s older brother, Ariel Quiroga, had a car accident in 2017, the lives of the five relatives turned upside down. The five months in a coma and the subsequent disability forced this humble Argentine family to focus all their efforts on getting him through. Mónica Barrera became the provider of the only salary that came into the house; her partner had to retire and both Maribel, the middle sister, and Facundo, the youngest, gradually stopped attending classes to sell cakes on the street and help at home. “I always thought it was going to be temporary, but what we earned went to medicines. I couldn’t afford his supplies or the transportation to school for the little ones,” says this 48-year-old mother via video call. The pandemic was the last setback. But the educational community and an algorithm created with artificial intelligence managed to keep the young man and 4,000 other students in the schools of the province of Mendoza.
The pandemic widened too many open wounds in Latin America. In the education sector—which was already battling high rates of school dropouts, technological backwardness, and a huge gap between rural and urban areas—many wondered what they would do with the extremely high levels of school dropouts when they saw that fewer and fewer students were connecting to virtual classrooms. In Argentina, 7.6% of high school students dropped out of school after Covid-19. This percentage was almost two percentage points higher in the province of Mendoza (9.1%). That is why the educational community decided to put an end to school dropouts; the beginning of many inequalities. And artificial intelligence became the great ally in this.
The education department, with support from CAF, the development bank of Latin America and the Caribbean, and a group of engineers from the University of Buenos Aires (UBA), created an algorithm to detect those at risk of dropping out. With this traffic light in hand, and with the names and surnames of the students, teachers and pupils, they set out to reverse these indicators with consultations and individualized attention. And even door to door. Thus, advisors from Colegio 4110 like Carolina Resca convinced dozens of families with contexts of vulnerability similar to the Quirogas. “We follow these students at higher risk closely,” she says. “We have managed to ensure that they do not feel alone and that the families understand that we can help them. We already knew that there was dropout, but many times we were late. This tool helped us to plan.”
Of the 8,402 students with medium and high risk alerts from first to fourth year of secondary school in Mendoza in 2023, more than half (4,236) continued their studies. “Our purpose is to use information and evidence to improve the efficiency of policies and address school dropout problems,” explains Cecilia Llambi, senior executive of CAF’s social development projects department. This organization invested 20,000 dollars to finance the generation of the artificial intelligence model and the training of provincial officials to use it. “It would be ideal to promote more initiatives like these, because they are needed. It is one of the great challenges of the continent,” adds Llambi.
The school found a way to support them through two measures: the creation of a school cafeteria that serves almost 180 daily free meals and a subsidy for the buses that transport the children. “That was what would be good for many students. We only finally saw it clearly with the data on the table,” explains Patricia Robles, director of Colegio 4110. In this Mendoza school, with a highly vulnerable context, the economy was often the main reason for dropping out of school. “Many of the children feel that they have to support them financially at home and that is why they do not continue with us. We had to be there for them,” she says.
The titanic collection of data showed that, at least in Mendoza, there are four clear reasons for dropping out: repeated absences —which led to the withdrawal of the status of regular student—, little family support, the accumulation of failed subjects and mental health problems. These causes did not vary much between boys and girls. However, Juan Kamienkowski, researcher at the Applied Artificial Intelligence Laboratory of the National Scientific and Technical Research Council (Conicet) and one of the project’s creators, would like to continue expanding the database and have more information on the socioeconomic situation in the homes, since this is a transversal factor in almost all the variables cited.
The importance of data
It all started with something key: data. Mendoza has been collecting endless information on 100% of students from preschool to higher education in a nominal information management system for over five years. Age, sex, background, parents’ jobs, internet access… Each child’s digital folder is growing year after year with a lot of comparable information from one to another. This is a real treasure for the creators of the algorithm. Or, as Kamienkowski says, “half the work has already been done.” “This allowed us to make predictions month by month because they had ordered and digitalized data,” says the professor from the Department of Computing at the UBA. “The first thing to know is that, to replicate this model, the provinces or communities have to make efforts to collect all this information. Without that, we do nothing.”
The algorithm, which during the first year showed a 92% accuracy in students at greatest risk of dropping out, will be fine-tuning its aim as the years go by, as it is trained with the real data of those who drop out of school. As Romina Durán, general director of Schools of Mendoza, explains, The system is a tool that predicts the level of risk: “Now, then, the educational institution is the one that works with each of the children.”
This institution accompanied the school guidance group and the psycho-pedagogical office of each educational center to establish guidelines and suggestions on a case-by-case basis. The most vulnerable schools, such as Facundo’s, received greater support and some of them received a financial bonus. “We are aware that in some areas we have to put more emphasis than in others,” says Durán. While 84.6% of young people from the highest income quintile in Latin America manage to graduate from secondary education, only 44.1% of those from the lowest income groups do so, according to Unicef.
The community makes the difference
Although Barrera always dreamed of seeing her son in class again, the final push was when Facundo’s teachers came to his house. “I felt very supported,” she says. “My son used to have to come to class with a bag of nylon and they helped us with the backpacks and uniforms and explained to me that they would subsidize our group. He felt special that they didn’t want to leave him behind. He is very excited.” This is, for both Barrera and Robles, the key to the success of the project, since all the tentacles of the educational sector were organized to launch an articulated intervention between educational centers, other students and families. “The most effective thing is that the children feel that they belong and that we all get involved,” concludes Durán.
This small but great revolution in Mendoza is, for many, the counterweight to a technology that has raised all the alarms. For Kamienkowski, this experience is a great example of how artificial intelligence is not only bringing infinite ethical questions, but also possibilities. “With this detection model we do not seek to stigmatize anyone. We are very responsible with the data and its confidentiality. But we have proven that it is a super powerful tool that can serve those who make decisions.”
#Artificial #intelligence #Facundo #students #school