Thomas Gammeltoft-Hansen receives Villum Synergy Grant of 6M DKK
The grant will fund the project: ”Explainable Artificial Intelligence and Credibility in Asylum Decision-making” (XAI-CRED) as it explores the use of explainable AI models in legal decision-making related to asylum cases.
The Villum Foundation has awarded Professor Thomas Moeslund of Aalborg University and Professor Thomas Gammeltoft-Hansen of UCPH a grant of collectively 12M DKK as part of the Villum Synergy Programme.
National asylum systems operate on the assumption that asylum should be granted only to those applicants in genuine need. However, making this determination in practice is complicated, as it involves complex, intersubjective assessments of the credibility of the applicant’s narrative. This has led to criticism that the existing systems are prone to bias and discrimination, resulting in what is often described as an “asylum lottery”. Applying AI to this area holds the potential for increased efficiency and fairness. At the same time, there are significant concerns that existing biases may be perpetuated, merely replacing the current human "black box" with a digital one.
This project seeks to address the “digital black box” problem by leveraging recent methodological advances from Explainable AI (XAI). Currently, XAI models primarily serve as 'debugging tools,' offering technical explanations for AI developers. What sets this project apart is its pioneering approach to a new type of XAI aimed directly at providing explanations that are relevant for domain experts. It will not only break new ground in terms of providing meaningful explanations to legal experts but also offer generalizable insights for basic research on human decision-making. As Thomas Gammeltoft-Hansen explains: “We wish to use AI to create more transparency around how we, as humans, make legal decisions, and thereby also reduce the risk of AI models blindly reproducing human bias.”
Read more about the Villum Synergy Programme here (In Danish)