As part of the National Police Lab AI (NPAI), I conduct research into making data-driven automated decision making (machine learning) transparent and explainable. This new research area, Explainable Artificial Intelligence (XAI), combines insights from algorithmic decision-making, philosophy, and (social) psychology.
For the Netherlands National Police, I support practitioners in embedding explainability and transparency during the design, development and operation of AI applications. My open-source Python toolkit Explabox helps data scientists to gain trust in their AI models, and to be confident in creating and applying AI models in a legally and ethically correct manner. Global Causal Analysis (GCA) depicts the effects of task-, fairness- and robustness-related features on global black-box AI behavior in a single graphical overview—fostering analysis and discussion.
In collaboration with ALGOPOL, I assess how (X)AI can meaningfully contribute to decision-making processes in the National Police, while taking into account end-user expectations and ethical principles.
In September 2019 my master's thesis Contrastive Explanation for Machine Learning won the Graduate School of Natural Sciences thesis award, and was runner-up for the UU Best Master's Thesis 2019.