Hundreds of researchers were challenged to predict the future of a group of children using a database collected over 15 years and machine learning methods. Despite the infinity of information and sophisticated technologies, no participants were able to make accurate predictions. In general, those predictions were practically the same as the ones obtained using simple methods, such as linear regression.
The study entitled “Fragile Families & Child Wellbeing Study”, developed by the Sociology Department at Princeton University, gathered data on almost 5,000 young people born in 2,000 cities with a population of over 200,000 in the United States.
In the proposed challenge, teams of social scientists looked at information collected at six moments in the children’s lives: at birth and when they were 1, 3, 5, 9, and 15 years old. From these data, they should predict the situation of children, parents, and housing in six different aspects, such as school performance, housing conditions, and risk of eviction, among others.
The result demonstrated that there is still progress to be made concerning predictive methods so that technology can make more elaborate projections. One fact that caught the attention of experts was that, despite using different methods, most teams made similar mistakes.
One of the conclusions was that the difficulty is not in the methods or limitations of machine learning but the set of information available. Despite having an enormous amount of data, it is not always possible to have access to some specific issues, such as the region in which they live or even the infrastructure available there.
The new mass collaboration model
The results raised questions about how far predictive ability can go for social aspects, even with a very large data set. They also make us reevaluate the limits of machine learning.
Even so, the study provides the basis for future research on the potential of technology to enable better living conditions. Another positive aspect is that all data and forecasts made are open source and can be used in analysis and research.
The challenge proposed by Princeton social scientists has also given rise to a new model of mass collaboration in the field of social sciences that can be the starting point for overcoming existing limitations and anticipating solutions to many of humanity’s problems.