top of page

Meet the Civic AI Lab Ph.D. Projects

In December last year, the Civic AI Lab was officially launched, with one goal in mind: develop AI systems that help advance society, while respecting fundamental human rights such as non-discrimination and equality of opportunity.

Towards that goal, five Ph.D. projects will be developed over the next four years, centered around the areas of education, health, well-being, environment, and mobility. In this blog post, we give an introduction to each of these projects and the candidates working on them.

Education Project

In this project we aim to create a machine learning algorithm that allows domain experts to define unwanted behavior while allocating financial resources for students based on their socio-economic background.

What can be unwanted behavior in a machine learning algorithm?

The financial resources allocated for students are currently determined based on variables like the education level of parents, socioeconomic and migration status. The fairness of this algorithm is unclear, creating a risk of unjust and unexpected bias.

How can we avoid unwanted behavior?

Seldonian algorithms offer a framework for creating machine learning algorithms that allow domain experts to define unwanted behavior without requiring additional data analysis. The goal of this project is to create a Seldonian algorithm that ensures a fair distribution of financial resources for students in the City of Amsterdam.

PhD candidate working on this

Mayesha has obtained a Bachelor’s degree in Computer Science and Master’s degree in Media Informatics. She has previously worked as a Web Developer and a Research Engineer specializing in Knowledge Graphs for conversational agents. Mayesha’s personal goal is to study and implement technologies that can accelerate positive growth for educational infrastructures. Towards that goal, she is now studying the safety and fairness of machine learning algorithms.

You can contact Mayesha via, on Linkedin, or Twitter.

Health project

In the health project, we will develop differentially fair classification and causal prediction machine learning models to help improve equal health access and treatment of patients in Amsterdam.

What is the health concern?

It is becoming increasingly clear that the first years of a child are essential in their development. This time influences life expectancy, the risk of illness, and their socioeconomic position later in life. However, not all children at the start of their lives have the same health opportunities due to a wide range of complex and interacting factors. A healthy start in life has an impact on people now and in the future.

How do we address the health concern?

Amsterdam has recently started the city-wide program “Gezonde en Kansrijke Start” (GKS). This program aims at “opportunity-reach and healthy start of life” and involves all relevant parties regarding pregnancy, birth, and early parenthood. The large amount of data collected offers the opportunity to connect different data sources and to extract new action perspectives using AI. This project will contribute to the GKS program by bringing in privacy-by-design and fairness-by-design principles into the data predictive modeling and analytics pipelines. We aim to mitigate the risks of harmful and unlawful health interventions and predictions, thus, improving healthcare in the city. Research results will be used to support professionals and guidelines in local, regional, and, possibly, national health care innovation.

PhD candidate working on this

Sara was born in Nicaragua and grew up in Guatemala. She holds a BSc in information systems engineering, an MBA, and an MSc in applied data science. Her diverse professional experience ranges from being an IT infrastructure administrator to teaching at university. Sara recently worked on Osteoarthritis research using machine learning techniques to identify phenotypes. She also volunteers for The Hague Peace Projects.

You can contact Sara via, on Linkedin, or Twitter.

Well-being project

In the well-being project we will develop explainable AI systems that help improve the well-being of the citizens of Amsterdam.

What is the well-being concern

The focus of this project is on children specifically, and on overweight and obesity in particular. This is because overweight and obesity in children leads to chronic well-being problems throughout life. Currently, almost 1 in 5 children in Amsterdam are overweight. That is almost one and a half times as much as the national average. In addition, it is known that this is an inequality issue in that it touches more children in families with a lower socioeconomic status and families with a non-western or a migration background. To improve on this, the City of Amsterdam started the Amsterdam Healthy Weight Program. The program’s mission is to let all children in the city grow up healthy by 2033.

How do we address the well-being concern

The Amsterdam Healthy Weight program is trying to achieve their objective through a Whole Systems Approach: this means that the focus is on an integrated approach with involvement of all relevant sectors and actors, such as parents and the children themselves, but also schools, sports organisations and food suppliers. This approach provides an opportunity: integrating data from all involved sectors and actors to yield new insights. These insights could offer new action perspectives, allowing for example to redirect resources for combating overweight and reducing inequalities with targeted health programs. In order to fully benefit from these insights and reliably assist health professionals and policy makers, we aim to develop explainable systems. Such systems should provide explanations that enable users to understand how the system reached its decisions and how the outcomes can be changed with the right interventions.

PhD candidate working on this

Ilse graduated from the interdisciplinary BSc Beta-Gamma and the MSc in Artificial Intelligence at the UvA. She has done internships in consultancy, computer vision and explainable AI and worked with an NGO advocating for an inclusive labour market. She has an interdisciplinary research interest and is eager to develop AI systems that empower people.

You can contact Ilse via

Environment Project

In the environment project, we look to create explainable vision models using panoramic imagery for the prediction of neighbourhood characteristics in Amsterdam.

Taking responsibility for black boxes

Measuring neighbourhood characteristics is a time consuming process when it is done through surveys of the population. Using deep vision models in combination with panoramic imagery for prediction of these characteristics has been shown to be a viable alternative. However, the decisions of these deep vision models are not interpretable by the people that will use them: civil or legal servants. Moreover, these models can be biased, as their predictions may lie in visual aspects that could be classified as unfair. As such, these methods can propagate inequality instead of tackling it.

How do we interpret black boxes

The main problem behind such a biased deep vision network, is that it’s a black box. The predictions it makes are not interpretable, and thus the users don’t know how or why a prediction has come to be. By developing vision models that are inherently more understandable we can allow lay users of the technology to understand the models and take responsibility for decisions based on its predictions.

PhD candidate working on this

Tim Alpherts completed a bachelor’s degree at Amsterdam University College in Computer Science and Mathematics, and a Master’s degree in Artificial Intelligence at the University of Amsterdam. He previously interned at the City of Amsterdam and worked on Visual Place Recognition.

You can contact Tim via or LinkedIn

Mobility project

In the mobility project, we aim to develop machine learning algorithms that extract mobility patterns, measure mobility poverty, and recommend flows to eliminate it.

What is mobility poverty

In today's fast-paced society, mobility is taken for granted. We assume that we can go from one place to the other quickly, cheaply, and without effort. However, not everyone has the privilege of such a service. Increasing rental prices, car usage becoming more and more expensive and a general shift of city planning towards utility instead of inclusivity put pressure on the mobility of vulnerable groups. As a consequence, some people are limited from participating in daily activities such as labor, education, and social events. This phenomenon is known as mobility poverty.

How do we address mobility poverty

There are two main challenges in tackling mobility poverty. Firstly, there is limited insight into the size and composition of the groups that suffer from it. In Amsterdam, we know that it mainly affects poor, non-western neighborhoods, but how many people and what their socio-economic factors are is not exactly known. The second challenge is recommending interventions that can lead to reducing mobility poverty. To overcome these challenges, we will use public mobility and socio-economic data to build machine learning models that extract, measure, and recommend poverty-reducing mobility flows.

PhD candidate working on this

Dimitris obtained his bachelor's degree in Computer Science at the Aristotle University of Thessaloniki and his master's degree in Data Science at the University of Amsterdam. Before joining the Civic-AI lab, he has worked as a software engineer, a web developer, and a data analyst. His research interests include tackling social inequalities, facilitating public service, and promoting data literacy.

You can contact Dimitris via, on Linkedin, or Twitter


bottom of page