Explainable Artificial Intelligence and Fairness in Asylum Law (XAIfair)

The scientific research approach in XAIfair follows three steps, each requiring interdisciplinary collaboration. First, an NLP machine learning algorithm aimed at predicting the outcome of an asylum decision from textual content is developed. This algorithm will serve as a testbed for the remaining project.

Second, an ‘indirect XAI algorithm’ is developed to serve as benchmark. Developing the indirect XAI algorithm requires close collaboration between data scientists and legal scholars both to ensure the development of a ‘common language’, and to enable an iterative process where algorithmic output will be assessed by domain experts that in turn provide feedback for further development, etc.

The third step will apply a neurosymbolic approach and develop a general XAI algorithm for asylum law. The approach will be rooted in interdisciplinary and ethical discussions regarding ‘what an XAI algorithm can explain’ and ‘what an XAI algorithm should explain’ to advance the ‘common language’ into an ontology.

The project is a partnership between the Nordic Asylum Law & Data Lab led by professor Thomas Gammeltoft-Hansen (iCourts, Faculty of Law, University of Copenhagen) and the AI for the People research center led by professor Thomas Moeslund (Faculty of IT and Design, Aalborg University).

Two international collaborators are connected to the project: Dr Nikolas Aletras (Department of Computer Science, University of Sheffield) and Søren Jørgensen (Center for Human Rights and International Justice, Stanford University)