I-SAFE: Wasserstein Coherence Metrics for Structural Auditing of Scientific AI Models

TL;DR

The paper introduces I-SAFE, a framework for auditing scientific AI models using Wasserstein coherence metrics to evaluate distributional alignment with domain knowledge, revealing differences invisible to accuracy metrics.

Contribution

It proposes a novel post-hoc distributional auditing framework, I-SAFE, utilizing Wasserstein coherence metrics to assess model outputs against structural priors.

Findings

I-SAFE reveals significant differences in model behavior not captured by accuracy.

The framework is applicable across domains with structured inputs and external priors.

Applied to drug-target interaction models, it uncovers distributional discrepancies.

Abstract

Deep learning models are increasingly used in scientific prediction tasks where strong benchmark performance is often interpreted as evidence of scientifically meaningful behavior. This interpretation is fragile, as models may exploit shortcut features, dataset-specific regularities, or distributional biases that are predictive on held-out data but not aligned with domain-relevant structure. To address this limitation, we introduce the \textsc{I-SAFE} (Interventional Secure, Accurate, Fair and Explainable) framework, a post-hoc distributional auditing framework for scientific AI models centered on the Wasserstein Coherence Metric (WCM). Given a trained black-box predictor and an external structural prior encoding domain knowledge about task-relevant input structure, \textsc{I-SAFE} evaluates raw model outputs under structurally guided perturbations of the input. The proposed audit measures output-distribution coherence through three complementary metrics: a Quantile-Based Metric (QBM) for location-level coherence, the WCM for ordinal coherence, and a translation-invariant WCM variant for shape coherence. We instantiate \textsc{I-SAFE} on drug--target interaction (DTI) prediction using the Davis kinase benchmark, KLIFS (Kinase--Ligand Interaction Fingerprints and Structures) binding-pocket annotations, and three sequence-based DTI models: DeepConvDTI, DeepDTA, and TAPB. Although the models operate in a comparable predictive regime, \textsc{I-SAFE} reveals substantially different distributional response profiles, a distinction invisible to accuracy-based evaluation. The framework is model-agnostic and applicable to any domain where inputs admit a structured decomposition and an external prior is available.