Business Ethics and Corporate Social Responsibility RSS

  Big data to help predict social exclusion 

Martínez de Albéniz, Victor; Echave Martínez, Cynthia
Print Share

  • Artificial intelligence (AI) techniques that have been used in banking, retail and other industries can also be put to work to support social inclusion efforts.

  • In Barcelona, age, building characteristics and participation in elections are among the relevant variables that help predict increases in social need.

  • The COVID-19 crisis is likely to require a retraining of the AI models with new data.

Photo: Kaspars Upmanis via Unsplash


With predictive models using artificial intelligence (AI) techniques and big data, professor Víctor Martínez de Albéniz is able to estimate social vulnerability in a territory. This is reflected in the research project "Big data at the service of social inclusion," funded by the RecerCaixa program, which identifies some key variables that public leaders should watch in order to promote social inclusion. IESE Insight talks with Martínez de Albéniz about the opportunities and challenges presented by AI in decision-making.

IESE Insight: First, why is big data helpful here?
Víctor Martínez de Albéniz: Big data creates lots of opportunities to make better decisions. I have spent many years developing big-data techniques based on artificial intelligence for the retail industry, to find out how customers behave at the point of sale. This project aims to help detect problems of social exclusion in Barcelona using the same sort of AI techniques to build predictive models. And similar approaches can be used to look at pollution or mobility in cities, for example.

II: What were the main findings of this research?
VMA: In collaboration with Barcelona's Department of Social Services, we collected these really large data sets -- the big data, if you will -- that include the needs of people in different parts of the city and also the factors that seemed to be driving those needs. We identified five families of drivers: (1) autonomy, which includes physiological factors, age and gender; (2) economic factors; (3) labor status; (4) environmental factors, such as connectivity and access to infrastructure; and (5) the identities of the different populations, including how many were born abroad or participated in elections. And we identified which variables seemed to lead people to seek out services or financial help -- and, in the case of financial aid, to what degree and for how long.

We found that the single biggest driver of dependency benefits -- for example, having a nurse visit regularly or someone to help with shopping -- was age. But asking for financial assistance was essentially driven by economic factors -- namely, having a low income and living amid high rental or real estate prices. We also found that living in older buildings and not participating in elections were among the indicators of social vulnerability.

With these sorts of models, we get a predictive algorithm that allows you to connect the patterns around these different drivers and how they evolve over time. And this is how your model can predict demand on social services. For example, if you see an indicator climbing much higher than it did in the past, the model will tell you the predicted impact on financial aid requests. We train the model with past data and then we use it to make better predictions.

II: Will the predictive models still be valid now that we're facing the COVID-19 crisis?
VMA: COVID-19 has triggered a macro crisis and requests for social aid have been climbing quite a bit. Because people's incomes have been hit, our model predicts that, but we will have to see if the model's predictions are accurate. I think that all the AI algorithms will probably need to be retrained in the future in order to take COVID-19 into account. That's true not only for our case, but in other areas as well. Our plan is to retrain the model with 2020 data and recalibrate our models for 2021.

The research project "Big data at the service of social inclusion" has been carried out from 2018 to 2020 with the support of RecerCaixa, a program promoted by the "la Caixa" Foundation, in collaboration with ACUP, the Catalan Association of Public Universities.
This article is based on:  Big Data at the service of social inclusion
Year:  2020
Language:  English