arXiv:2606. 18479v1 Announce Type: new Abstract: Reject inference methods are widely used to mitigate survival bias in credit scoring, yet their effectiveness remains poorly understood.
By Bruno Scarone, Ricardo Baeza-Yates
The paper introduces a causal mediation framework to separate direct discrimination from structural inequality in AI-driven credit decisions. Using Pearl’s natural direct and indirect effects, it presents an identification strategy under treatment‑induced confounding and proposes a doubly‑robust estimator with efficiency guarantees. Empirical analysis of 89,465 mortgage applications shows that about 77% of racial denial disparities stem from financial mediators, while the remaining 23% represents a conservative lower bound on direct discrimination.
By Duraimurugan Rajamanickam
arXiv:2607. 19526v1 Announce Type: new Abstract: "Stop Chasing the C-index when Evaluating Survival Analysis Models" (ICML 2026, Spotlight) argued normatively, on synthetic data, that evaluating survival models by discrimination alone, i.
By Rafael da Silva, Danilo Alvares
arXiv:2609.27654v1 Announce Type: cross
Abstract: Verified income is often unavailable in digital loan applications, forcing lenders to rely on reported income and potentially leading to over-lending...
By Sultan Amed, Tanmay Sen, Sayantan Banerjee
arXiv:2504. 21259v2 Announce Type: replace-cross Abstract: Accurate imputation of race and ethnicity (R&E) is essential for fair lending compliance under ECOA, HMDA, and the Community Reinvestment Act, where up to 15% of mortgage applications carry missing race data and regulated institutions bear responsibility for identifying disparities on those records.
By S. Chalavadi, A. Pastor, T. Leitch
The paper presents a locked, multi‑signal audit protocol designed to detect supervision drift in credit‑risk models that use proxy labels. It comprises five layers—transfer performance, an oracle‑gap probe, a calibration diagnostic, feature‑label stability, and a synthetic positive control—each with predefined thresholds and decision rules. Applied to a public LendingClub dataset, the protocol shows stable ranking, small oracle gaps, and identifies a prevalence and probability‑scale mismatch that recalibration largely mitigates, though its root cause remains unclear.
By Mehrdad Shoeibi, Muhammad Shabanpour, Waldemar Karwowski, Niloofar Yousefi