Model-Driven Decisions Lab (MoDDL) 
Collaborative project with the National Police

In modern-day policing, machine learning, causal models, and simulations for decision-making are crucial. Harnessing such models is not only about analysing and interpreting massive amounts of data, it’s also about human-AI interaction, ethics and privacy. With the Model-Driven Decisions Lab (MoDDL), the Netherlands Police and TU Delft are taking the next step to increase insights into the digital world and make better use of them. This collaboration also explores other approaches to police work for improvements in effectiveness, in a responsible way.

The lab is geared to contributing to the various aspects of data and model-driven decision-making in the security domain. The five unique research lines harness a combination of technology and science, and values such as accountability, explainability and transparency, which play a key role in decision-making. Topics range from AI-assisted long-term decision-making, creating situational awareness using sensors and edge computing, and new machine learning methods based on sequential decision-making, to causal models for real-time intelligence and AI-assisted intervention.

Projects within the MoDDL Lab

Five PhD students will work together within MoDDL in the coming years. Four of them presented their project and further plans during the opening. Together with stakeholders from the Netherlands Police and scientists from TU Delft, the projects were defined. The Safety & Security Institute and the AI innovation team also played an important role in this.

Coen van den Elshout will focus on dynamic modelling of criminal power structures. By working at MoDDL, he can learn from real-life experience. In addition, he can use system dynamics to generate social impact. He sees these as major advantages. Coen is supervised by Floortje d’Hont, Willem Auping and Jan Kwakkel.

Tess Jongsma was hired as a PhD to investigate the impact of demographic and societal changes on police organisation. Her background in criminology and system dynamics makes this research project entirely her cup of tea. She is interested in investigating a social phenomenon, such as the interaction between crime and mental health. Tess is supervised by Floortje d’Hont, Willem Auping and Jan Kwakkel.

As PhD, Pascal Benschop will investigate automated activity detection from multiple sensors and the impact on privacy. His passion for computing and data vision made him excited for this PhD position. He sees the potential and need for this research in the police and hopes to support the police with his application. Pascal is supervised by Justin Dauwels and Jan van Gemert.

Andreea Minculescu joins MoDDL to work on Reinforcement Learning. A subject that suits her well due to her background in Artificial Intelligence. The MoDDL Lab appealed to her because of the opportunity to apply her research more practically. She finds the potential and responsibilities of research a great opportunity to seize. Andreaa is supervised by Matthijs Spaan and Frans Oliehoek.