AI for Social Good & Equality of Opportunity

On 25 September 2015 at the annual UN meeting, 193 countries signed an agreement on the 17 Sustainable Development Goals (SDG) and 169 associated targets. These goals aim to move us towards an inclusive, just and sustainable society by 2030. The first goal is to end extreme poverty. Other goals focus on health, education, sustainable energy, climate change and reducing inequality. Over the past five years, some hopeful results have been achieved according to the SDG 2020 report. For example, the rate of children and youth dropping out of school has decreased; access to safely managed drinking water has improved; and women’s representation in leadership roles has increased. At the same time, there’s still an enormous amount of work to be done in the next ten years: the number of people suffering from food insecurity is on the rise, the natural environment continues to deteriorate at an alarming rate, and dramatic levels of inequality persist throughout the world.

Can AI help achieve the SDG’s?

In a recent study published in Nature Communications, Vineusa et al. showed that AI has a positive impact on 80% of the SDGs.  AI, for example, enables the creation of circular economies and smart cities that efficiently use their resource and underpin low-carbon systems. However, AI also seems to increase inequality in many socio-economic areas. In part this is because AI is often used with the aim of optimising speed and minimising costs, rather than promoting equality of opportunity.  A good example is the recent controversy in the UK, where an algorithm was used to predict final exam results for students who were unable to take their final exams because of the COVID-19 pandemic. This was thought to be a quick and cost effective way to assign grades to students. However, because the algorithm was calibrated using location data instead of relying on individual performance alone, it produced biased results for students from less advantaged socioeconomic backgrounds.

The Civic AI Lab

It is against this background of deploying AI for good that we launched the 15th ICAI lab this summer: Civic AI Lab (CAIL). CAIL is a collaboration between the City of Amsterdam, the Ministry of Interior Affairs, Vrije Universiteit Amsterdam (VU) and the University of Amsterdam (UvA). The lab’s vision is a society in which citizens from all backgrounds have equal opportunity to contribute to and benefit from an innovative and thriving society. The lab’s mission is to develop AI technology that promotes economic and social human rights, such as the right to health, education and employment, while respecting fundamental human rights such as non-discrimination and equality. We will start with five PhD projects in the areas of education, environment, mobility, health and well-being. The projects will be carried out within interdisciplinary teams of (AI) scientists, and on the basis of use cases and data provided by the City of Amsterdam. Each project aims to create insights into inequality of opportunity in the City of Amsterdam and to create new ways to increase equality of opportunity with the help of new AI technologies. For example, in the education project we work together with the city’s education department to create fairer models for the distribution of finances between schools in order to give school children in Amsterdam the fairest possible chance of a good education.

A call to action

In September 2020, on the fifth anniversary of the SDGs, dozens of Dutch organizations, including all universities held a flag campaign to express their support for the United Nations’ Sustainable Development Goals. We are using this anniversary as an opportunity to call on AI scientists and Data Science researchers in Amsterdam and beyond to step in, and step up efforts for equal opportunity and SDGs in general. We all need to shift up a gear, now that COVID-19 is leading to an unprecedented health, economic and social crisis and making the achievement of SDGs even more challenging.

Why we need the UvA Data Science Centre

Accelerating science through data

As scientists and researchers, we are already immersed in what Turing Award winner Jim Gray termed the 4th paradigm: the use of data-intensive techniques to attack the hardest problems in research. This can clearly be seen in how researchers have tackled the many, sometimes unexpected, problems that have arisen from the COVID-19 pandemic. For example, what is the real-time economic impact of COVID-19? Through the use of novel data sources and data science techniques, Raj Chetty and his team at Harvard have been answering that question. During the Amsterdam Data Science meetup series on COVID-19, we saw how data science is being used to answer diverse research questions, such as understanding the virus’s impact on our transportation system or building better predictive models for the disease. At the University of Amsterdam (UvA), rarely a week goes by where I don’t see my colleagues in a variety of disciplines engaging with data to do impactful science. Recent examples include:

The driving forces behind the Data Science Centre

Given these compelling examples of data-driven research already taking place at the university, we asked ourselves how can we further support and expand the use of such techniques in all research domains across the UvA? In our discussions with researchers and faculties, one of the clear needs was for skilled data scientists and engineers to be part of research teams. We see this from Chetty’s experience, where the availability of data scientists and engineers as part of the team was central to being able to answer the research question. They were able to develop the required data pipelines, integrate heterogeneous datasets and anonymize data correctly. Personally, it’s clear to me that this team science oriented approach is critical for data-driven research.  The second area where we saw opportunities for acceleration was interdisciplinary cooperation oriented around data. We know that from large scale studies that communities arise around datasets and that these communities often draw from different disciplines and methodological backgrounds. For example, in the AI4Science lab, researchers who study methods of causal discovery are working with those who work on gene regulatory networks to further research in both areas. In the case of studying urban environments mentioned above, urban geographers and political scientists came together around street view datasets.  To address these needs, the UvA will launch a Data Science Centre (DSC) in 2021. The centre was inspired by similar initiatives at NYU, Berkley and more locally by the Netherlands e-Science Centre and the University of Maastricht. However, we aim to have our own unique flavour that will leverage the UvA’s strengths.

Data Science Centre activities

First, the DSC will embed data scientists and engineers throughout every faculty in the university. We will create a community where knowledge sharing is key. Data scientists and engineers will share knowledge and experience through weekly sessions at the Library. The Library has always served as an intellectual crossroads within universities where scientists acquire knowledge and skills making it a natural home for the DSC. We aim to have 35 dedicated staff by 2025 and will offer training opportunities for faculty and staff members more broadly. This will ensure that faculty members will be able to act on new skills gained and the ideas and knowledge sharing happening in the DSC. Second, following the success of collaborative interdisciplinary PhDs and researcher assistants (as seen in the UvA’s research priority areas, AI4 Science and ICAI Amsterdam labs), the DSC will adopt a similar program to innovate data science methods through engagement with hard domain specific research questions.

The Data Science Ecosystem

Just as the UvA helped spur the Amsterdam Data Science (ADS) ecosystem over five years ago, it is now harnessing the ecosystem to improve its broader research. The DSC will work together with ADS to facilitate data science related collaboration and networking in and outside the university. Indeed, this last point shouldn’t be underestimated: data driven research often requires collaboration with parties outside the research domain. Again, reflecting on the COVID-19 economy-tracking example, the data used is supplied by multiple commercial parties. Thus, connecting to the wider data science community through ADS is critical for enabling world-class research. Using the strengths of the UvA and ADS ecosystem, I’m excited to see the new discoveries that the UvA Data Science Centre will make possible. If you are interested in collaborating with the centre, please contact