Marcel Robeer
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Marcel Robeer has a total of 16 co-authors. Here are the top co-authors by number of shared publications:
- Floris Bex
Utrecht University
3 shared publications - Elize Herrewijnen
National Police Lab AI, Utrecht University, The Netherlands
2 shared publications - Michiel Pieter Bron
Utrecht University
2 shared publications Riwish Hoeseni
Netherlands National Police, The Netherlands
2 shared publications- Ad J. Feelders
Utrecht University
1 shared publication Fabiano Dalpiaz
1 shared publication
Latest Publications
- 28 October 2025
- Published by Utrecht University Library
Beyond Trust: A Causal Approach to Explainable AI in Law Enforcement
An AI can process a million clues in a second. But what if its most critical clue is just a hidden bias? The increased integration of opaque Artificial Intelligence (AI) systems into high-stakes domains, such as law enforcement, presents critical challenges to accountability, fairness, and public trust. While Explainable AI (XAI) aims to mitigate this "black-box" problem, many current methods provide superficial, correlational insights that are insufficient for environments demanding verifiable understanding. This dissertation argues for a fundamental shift Beyond Trust: a move from accepting superficial explanations to demanding rigorous, verifiable insights. From using unevaluated methods to requiring robust validation, and from developing technical tools to ensuring they are operationalized for practitioners. This work contributes to this shift by developing and validating a suite of causality-grounded techniques and context-aware tools designed to foster genuinely understandable and accountable AI. The research is structured around three core challenges: (1) the lack of contextual awareness in XAI development, (2) the limited explanatory power and evaluation of current techniques, and (3) significant usability barriers for practitioners. The dissertation first establishes a set of guiding desiderata for XAI in law enforcement (Chapter 2), emphasizing versatility, an appropriate focus on NLP, and a grounding in causal reasoning. The dissertation's primary technical contributions (Chapters 3-5) address the challenge of explanatory power by introducing three novel artifacts. First, CounterfactualGAN generates realistic and plausible local counterfactual explanations ("what-if" scenarios) for NLP models, validated through both functional metrics (including a novel perceptibility score) and human-grounded evaluations of naturalness. Second, Global Causal Analysis (GCA) constructs an explanatory causal graph of a model's global behavior, uniquely inferring high-level features from raw text. It is validated via a novel three-step framework including Z-fidelity metrics and expert-driven sanity checks. Third, the Causality for XAI (C4X) Framework provides a conceptual guide for the rigorous application of Causal Effect Estimation (CEE) in XAI. It contributes a structured process to ensure transparency and prevent the misapplication of causal methods, serving as a roadmap for researchers to apply rigorous, multi-faceted evaluations to their own causal explanations. To bridge the gap between research and practice, the Explabox (Chapter 6) operationalizes these principles into an open-source, user-friendly software toolkit for NLP. It integrates XAI, fairness, robustness, and security analyses into a structured audit workflow. Its utility was validated through an application-grounded design process with the Netherlands National Police. The dissertation concludes (Chapter 7) that these contributions successfully advance the Beyond Trust paradigm, discussing the work's theoretical, practical, and societal implications, as well as limitations and an agenda for future research.
- 28 October 2025
- Published by Utrecht University Library
Beyond Trust: A Causal Approach to Explainable AI in Law Enforcement
An AI can process a million clues in a second. But what if its most critical clue is just a hidden bias? The increased integration of opaque Artificial Intelligence (AI) systems into high-stakes domains, such as law enforcement, presents critical challenges to accountability, fairness, and public trust. While Explainable AI (XAI) aims to mitigate this "black-box" problem, many current methods provide superficial, correlational insights that are insufficient for environments demanding verifiable understanding. This dissertation argues for a fundamental shift Beyond Trust: a move from accepting superficial explanations to demanding rigorous, verifiable insights. From using unevaluated methods to requiring robust validation, and from developing technical tools to ensuring they are operationalized for practitioners. This work contributes to this shift by developing and validating a suite of causality-grounded techniques and context-aware tools designed to foster genuinely understandable and accountable AI. The research is structured around three core challenges: (1) the lack of contextual awareness in XAI development, (2) the limited explanatory power and evaluation of current techniques, and (3) significant usability barriers for practitioners. The dissertation first establishes a set of guiding desiderata for XAI in law enforcement (Chapter 2), emphasizing versatility, an appropriate focus on NLP, and a grounding in causal reasoning. The dissertation's primary technical contributions (Chapters 3-5) address the challenge of explanatory power by introducing three novel artifacts. First, CounterfactualGAN generates realistic and plausible local counterfactual explanations ("what-if" scenarios) for NLP models, validated through both functional metrics (including a novel perceptibility score) and human-grounded evaluations of naturalness. Second, Global Causal Analysis (GCA) constructs an explanatory causal graph of a model's global behavior, uniquely inferring high-level features from raw text. It is validated via a novel three-step framework including Z-fidelity metrics and expert-driven sanity checks. Third, the Causality for XAI (C4X) Framework provides a conceptual guide for the rigorous application of Causal Effect Estimation (CEE) in XAI. It contributes a structured process to ensure transparency and prevent the misapplication of causal methods, serving as a roadmap for researchers to apply rigorous, multi-faceted evaluations to their own causal explanations. To bridge the gap between research and practice, the Explabox (Chapter 6) operationalizes these principles into an open-source, user-friendly software toolkit for NLP. It integrates XAI, fairness, robustness, and security analyses into a structured audit workflow. Its utility was validated through an application-grounded design process with the Netherlands National Police. The dissertation concludes (Chapter 7) that these contributions successfully advance the Beyond Trust paradigm, discussing the work's theoretical, practical, and societal implications, as well as limitations and an agenda for future research.
- Open Access
- 13 October 2025
- Published by The Open Journal in Journal of Open Source Software
- Vol. 10 (114) , 8253
Explabox: A Python Toolkit for Standardized Auditing and Explanation of Text Models
Developed to meet the practical machine learning (ML) auditing requirements of the Netherlands National Police, Explabox is an open-source Python toolkit that implements a standardized four-step analysis workflow: explore, examine, explain, and expose. The framework transforms models and data (ingestibles) into interpretable reports and visualizations (digestibles), covering everything from data statistics and performance metrics to local and global explanations, and sensitivity testing for fairness, robustness, and security. Designed for developers, testers, and auditors, Explabox operationalizes the entire audit lifecycle in a reproducible manner. The initial release is focused on text classification and regression models, with plans for future expansion.
- 13 October 2025
- Published by The Open Journal in Journal of Open Source Software
- Vol. 10 (114) , 8253
Open AccessExplabox: A Python Toolkit for Standardized Auditing and Explanation of Text Models
Developed to meet the practical machine learning (ML) auditing requirements of the Netherlands National Police, Explabox is an open-source Python toolkit that implements a standardized four-step analysis workflow: explore, examine, explain, and expose. The framework transforms models and data (ingestibles) into interpretable reports and visualizations (digestibles), covering everything from data statistics and performance metrics to local and global explanations, and sensitivity testing for fairness, robustness, and security. Designed for developers, testers, and auditors, Explabox operationalizes the entire audit lifecycle in a reproducible manner. The initial release is focused on text classification and regression models, with plans for future expansion.
- Open Access
- 27 March 2025
- Published by Springer Nature
- p. 61-82
Explainable AI Is No Silver Bullet: Towards a Contextual Understanding of Appropriate Reliance on AI in Law Enforcement
Police officers increasingly rely on AI recommendations in their decision-making processes. This chapter delves into the potential and challenges of AI in law enforcement, focusing on the role of Explainable AI (XAI) on appropriate reliance, where appropriate is neither stubbornly rejecting, nor complacently accepting AI recommendations. We discuss recent XAI research findings and their implications for law enforcement, contrasting the limitations of both generalistic human-centred lab studies and application-grounded studies. We explore how XAI can impact the law enforcement environment, drawing from organisational and sociological literature, and from empirical examples from our research at the Netherlands Police Lab for Artificial Intelligence. Our insights show that XAI is no silver bullet towards appropriate reliance but—when designed well—it is one of the factors in the context of a decision that pushes or pulls a person towards appropriate reliance. Ideally, XAI provides considerations to decide to follow, reject, or cross-check an AI’s recommendation. We propose a contextual model to understand when and how XAI, along with other contextual factors, can facilitate appropriate reliance on AI recommendations. This model can help to identify individual and task characteristics to assess what factors foster the appropriate reliance on AI recommendations in specific organisational settings.
- 27 March 2025
- Published by Springer Nature
- p. 61-82
Open AccessExplainable AI Is No Silver Bullet: Towards a Contextual Understanding of Appropriate Reliance on AI in Law Enforcement
Police officers increasingly rely on AI recommendations in their decision-making processes. This chapter delves into the potential and challenges of AI in law enforcement, focusing on the role of Explainable AI (XAI) on appropriate reliance, where appropriate is neither stubbornly rejecting, nor complacently accepting AI recommendations. We discuss recent XAI research findings and their implications for law enforcement, contrasting the limitations of both generalistic human-centred lab studies and application-grounded studies. We explore how XAI can impact the law enforcement environment, drawing from organisational and sociological literature, and from empirical examples from our research at the Netherlands Police Lab for Artificial Intelligence. Our insights show that XAI is no silver bullet towards appropriate reliance but—when designed well—it is one of the factors in the context of a decision that pushes or pulls a person towards appropriate reliance. Ideally, XAI provides considerations to decide to follow, reject, or cross-check an AI’s recommendation. We propose a contextual model to understand when and how XAI, along with other contextual factors, can facilitate appropriate reliance on AI recommendations. This model can help to identify individual and task characteristics to assess what factors foster the appropriate reliance on AI recommendations in specific organisational settings.
- Open Access
The Explabox: Model-Agnostic Machine Learning Transparency & Analysis
We present the Explabox: an open-source toolkit for transparent and responsible machine learning (ML) model development and usage. Explabox aids in achieving explainable, fair and robust models by employing a four-step strategy: explore, examine, explain and expose. These steps offer model-agnostic analyses that transform complex 'ingestibles' (models and data) into interpretable 'digestibles'. The toolkit encompasses digestibles for descriptive statistics, performance metrics, model behavior explanations (local and global), and robustness, security, and fairness assessments. Implemented in Python, Explabox supports multiple interaction modes and builds on open-source packages. It empowers model developers and testers to operationalize explainability, fairness, auditability, and security. The initial release focuses on text data and models, with plans for expansion. Explabox's code and documentation are available open-source at https://explabox.readthedocs.io/.
Open AccessThe Explabox: Model-Agnostic Machine Learning Transparency & Analysis
We present the Explabox: an open-source toolkit for transparent and responsible machine learning (ML) model development and usage. Explabox aids in achieving explainable, fair and robust models by employing a four-step strategy: explore, examine, explain and expose. These steps offer model-agnostic analyses that transform complex 'ingestibles' (models and data) into interpretable 'digestibles'. The toolkit encompasses digestibles for descriptive statistics, performance metrics, model behavior explanations (local and global), and robustness, security, and fairness assessments. Implemented in Python, Explabox supports multiple interaction modes and builds on open-source packages. It empowers model developers and testers to operationalize explainability, fairness, auditability, and security. The initial release focuses on text data and models, with plans for expansion. Explabox's code and documentation are available open-source at https://explabox.readthedocs.io/.
- 30 October 2023
- Published by Springer Nature
- p. 299-323
Explaining Model Behavior with Global Causal Analysis
- 30 October 2023
- Published by Springer Nature
- p. 299-323
Explaining Model Behavior with Global Causal Analysis
- Open Access
- 20 May 2017
- Published by Springer Nature in Requirements Engineering
- Vol. 22 (3) , 339-358
Extracting conceptual models from user stories with Visual Narrator
- 20 May 2017
- Published by Springer Nature in Requirements Engineering
- Vol. 22 (3) , 339-358
Open AccessExtracting conceptual models from user stories with Visual Narrator
- 1 September 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 196-205
Automated Extraction of Conceptual Models from User Stories via NLP
Natural language (NL) is still the predominant notation that practitioners use to represent software requirements. Albeit easy to read, NL does not readily highlight key concepts and relationships such as dependencies and conflicts. This contrasts with the inherent capability of graphical conceptual models to visualize a given domain in a holistic fashion. In this paper, we propose to automatically derive conceptual models from a concise and widely adopted NL notation for requirements: user stories. Due to their simplicity, we hypothesize that our approach can improve on the low accuracy of previous works. We present an algorithm that combines state-of-the-art heuristics and that is implemented in our Visual Narrator tool. We evaluate our work on two case studies wherein we obtained promising precision and recall results (between 80% and 92%). The creators of the user stories perceived the generated models as a useful artifact to communicate and discuss the requirements, especially for team members who are not yet familiar with the project.
- 1 September 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 196-205
Automated Extraction of Conceptual Models from User Stories via NLP
Natural language (NL) is still the predominant notation that practitioners use to represent software requirements. Albeit easy to read, NL does not readily highlight key concepts and relationships such as dependencies and conflicts. This contrasts with the inherent capability of graphical conceptual models to visualize a given domain in a holistic fashion. In this paper, we propose to automatically derive conceptual models from a concise and widely adopted NL notation for requirements: user stories. Due to their simplicity, we hypothesize that our approach can improve on the low accuracy of previous works. We present an algorithm that combines state-of-the-art heuristics and that is implemented in our Visual Narrator tool. We evaluate our work on two case studies wherein we obtained promising precision and recall results (between 80% and 92%). The creators of the user stories perceived the generated models as a useful artifact to communicate and discuss the requirements, especially for team members who are not yet familiar with the project.