Explainability and the Epistemic Division of Labour in Adjudication

Abstract

The ‘black box’ quality of contemporary algorithmic tools raises concerns related to their use in court because of the law’s emphasis on explanations, transparency, and public reasons. We argue that the problems of explainability associated with contemporary algorithmic tools are, from a legal perspective, neither sui generis nor irreconcilable with existing norms. We distinguish between the types of explanations required by fact-finders and those required from judges. We conclude that apparent tensions can be reconciled by attending to the epistemic division of labour between the legal and scientific communities, contextualizing expert evidence appropriately, and distinguishing between explanation as reconstruction and as justification.