Asylum Decision Research Infrastructure (ADRI): Developing AI and metadata to establish and provide access to a comprehensive research corpus for analyzing Swedish Migration Agency decisions

The right to asylum is one of the most debated issues of our time.

As the first instance in the asylum process, the Swedish Migration Agency plays a central role in assessing asylum applications. The agency acts on behalf of the public and enforces the legislator’s intentions.

Its decisions are crucial for understanding the dynamics of the asylum process, the mechanisms of decision-making, and the impacts of migration policy.

In the Nordforsk-funded project Nordic Refugee Determination: Advancing Data Science in Migration Law (NordASIL), researchers at Uppsala University have gained access to about 590,000 digitized decisions from the Migration Agency’s asylum adjudication units between 2006 and 2022.

These decisions have been compiled in PDF format in Uppsala University’s data storage system for large-scale analysis of sensitive personal data.

The collection represents a significant dataset with major potential for interdisciplinary and multidisciplinary research on asylum-related migration.However, the research utility of the collection is currently limited due to its lack of metadata structure. Metadata that contextualizes and describes the data is essential for effective analysis.

Without a uniform structure, it is difficult to manage and utilize the dataset optimally.

The project aims, within a collaboration between the disciplines of law, statistics, social medicine, and linguistics, to enhance and expand the usability of the collection by creating a well-organized research corpus with annotated metadata.

Annotation involves processing the texts and providing them with detailed descriptions and labels, which facilitates identification and analysis.The project has three main goals:To develop a comprehensive research corpus with annotated metadata for qualitative and quantitative analysis of Migration Agency decisions between 2006 and 2022, with potential to include decisions from later years and expand metadata.To use machine learning to automatically identify and categorize linguistic expressions related to demographics, outcome, case type, permit ground, evidence evaluation, and medical evidence.To promote broader use of the material for research purposes.By developing AI and metadata to create and provide access to a comprehensive research corpus for analysing Migration Agency decisions, the project aims to strengthen Swedish migration research and establish international standards for comparative studies in asylum-related migration.