Signal Against Explainable AI in Law: Why Justifiable AI Matters (Credit Scoring Example)

Summary

This study reviews the EU's legal approach to AI explainability requirements through the lens of credit scoring. The author argues that the narrow technical reading of 'explainable AI,' or XAI, that currently prevails is poorly suited to legal contexts, and proposes 'justifiable AI' as a more appropriate standard. The paper examines the relevant EU legal provisions and reinterprets them by drawing on insights from technical science. It contends that the growing sophistication of machine-learning models used in credit scoring is deepening explainability problems rather than resolving them. The paper's title itself frames this argument concisely, casting justifiable AI as the standard that matters more than explainable AI in law, illustrated through the credit-scoring example.