An Explainable Hybrid AI Approach to Case Based and Rule Based Legal Reasoning
A rule-based model of the statute controls when a language model may act, and the caseworker makes the findings. With colleagues at Copenhagen and Bergen.
computer science and law · University of Copenhagen
I am a postdoctoral researcher at the University of Copenhagen, Department of Computer Science, in the Software, Data, People & Society section. In the LEXplain project, with the Faculties of Law at Copenhagen and Bergen, I study how AI-assisted decisions in welfare administration can meet the legal duty to give reasons: what a system must record for a decision to be explained, justified and contested by the person it affects.
My PhD at the University of Amsterdam (2026) showed that the legal norms governing health-data sharing, GDPR duties and data sharing agreements, can be formally specified and automatically enforced in access control. The work was evaluated on an international registry of paediatric brain tumours, with the Princess Máxima Center for paediatric oncology. I was supervised by Prof. Tom van Engers, Dr. L. Thomas van Binsbergen, and Dr. Dannis van Vuurden.
Before the PhD I trained in security and privacy (double MSc, Saarland University and University of Trento) and worked as a security analyst. I hold a BSc in Computer Science from Jimma University in Ethiopia.
A rule-based model of the statute controls when a language model may act, and the caseworker makes the findings. With colleagues at Copenhagen and Bergen.
The reasons in a decision letter are derived by rule from what was found in the case, and a language model only words them.
Milen G. Kebede (PhD thesis, University of Amsterdam, 2026)
L. Thomas van Binsbergen, Marten C. Steketee, Milen G. Kebede, Heleen L. Janssen, Tom M. van Engers (arXiv preprint, 2025)
Jamila Alsayed Kassem, Tim Müller, Christopher A. Esterhuyse, Milen G. Kebede, Anwar Osseyran, Paola Grosso (Future Generation Computer Systems 165, 2025)
Jamila Alsayed Kassem, Corinne Allaart, Saba Amiri, Milen G. Kebede, Tim Müller, Rosanne Turner, and others (Commit2Data, Schloss Dagstuhl, 2024)
Milen G. Kebede, L. Thomas van Binsbergen, Tom M. van Engers, Dannis G. van Vuurden (Legal Knowledge and Information Systems (JURIX), 2023)
L. Thomas van Binsbergen, Milen G. Kebede, Joshua Baugh, Tom M. van Engers, Dannis G. van Vuurden (Procedia Computer Science 198, 2022)
Milen G. Kebede, Giovanni Sileno, Tom M. van Engers (AI Approaches to the Complexity of Legal Systems (AICOL), Springer, 2021)
Milen G. Kebede (AI Approaches to the Complexity of Legal Systems (AICOL), Springer, 2021)
From 2024 to 2026 I was a Junior Lecturer in the Information Studies master's programme at the University of Amsterdam. I ran the lab sessions for Human–AI Interaction and supervised semester-long team projects in the Data Systems Project course, several of them on legal AI.
I supervised twelve theses, on topics including privacy notices and legal design, consent management, access control, and language models for regulatory and clinical information.
I was also a teaching assistant in Policy Making & Rule Governance, Intelligent Interactive Systems, Information Organisation, Modelling System Dynamics, and Societal Complexity and Designing with Data.