TaylorPODA: A Taylor Expansion-Based Method to Improve Post-Hoc Attributions for Opaque Models

TL;DR

TaylorPODA introduces a Taylor expansion-based framework with postulates to improve the systematic quantification of feature contributions in post-hoc explanations for opaque models, enhancing trustworthiness and interpretability.

Contribution

It proposes a novel TaylorPODA method guided by new postulates, including an adaptation property, to produce more principled and task-aligned feature attributions.

Findings

Achieves competitive attribution results against baselines

Provides explanations with stronger theoretical grounding

Enhances visualization-friendly interpretability

Abstract

Existing post-hoc model-agnostic methods generate external explanations for opaque models, primarily by locally attributing the model output to its input features. However, they often lack an explicit and systematic framework for quantifying the contribution of individual features. Building on the Taylor expansion framework introduced by Deng et al. (2024) to unify existing local attribution methods, we propose a rigorous set of postulates -- "precision", "federation", and "zero-discrepancy" -- to govern Taylor term-specific attribution. Guided by these postulates, we introduce TaylorPODA (Taylor expansion-derived imPortance-Order aDapted Attribution), which incorporates an additional "adaptation" property. This property enables alignment with task-specific goals, especially in post-hoc settings lacking ground-truth explanations. Empirical evaluations demonstrate that TaylorPODA achieves competitive results against baseline methods, providing principled and visualization-friendly explanations. This work enhances the trustworthy deployment of opaque models by offering explanations with stronger theoretical grounding.