Explainable Artificial Intelligence and Credibility in Asylum Decision-making (XAI-CRED)

AI technology has recently shown impressive capabilities and is being considered for public use across a broad range of sectors – including in highly sensitive areas such as asylum law. National asylum systems rests on the assumption that asylum is granted only to those in need and that such a need is known to decision-makers. What sets XAI-CRED apart is the pioneering of a new type of XAI aimed directly towards explanations relevant for domain experts. This will involve the notion of computational meaningful concepts that integrate both formal and practical domain knowledge into our XAI model. Moreover, XAI-CRED is the first project specifically focusing on credibility as what needs to be explained in XAI and law. This will break new ground in terms of providing meaningful explanations to legal experts, but also provide generalisable insights for basic research on human decision-making. Finally, XAI-CRED moves beyond existing debates by arguing that effective XAI models not only answer regulatory demands in the context of automatisation, but also represent an important opportunity for feeding back and enriching human decision-making.