Priority-Aware Shapley Value

TL;DR

The paper introduces Priority-Aware Shapley Value (PASV), a novel method for data valuation and feature attribution that accounts for contributor dependencies and priorities, improving fairness and interpretability.

Contribution

It proposes PASV, a flexible extension of Shapley values that incorporates precedence constraints and priority weights, with a scalable Monte Carlo estimation method.

Findings

PASV provides more structure-faithful allocations in experiments.

The method effectively incorporates dependency and priority information.

Experiments demonstrate improved interpretability and sensitivity analysis.

Abstract

Shapley values are widely used for model-agnostic data valuation and feature attribution, yet they implicitly assume contributors are interchangeable. This can be problematic when contributors are dependent (e.g., reused/augmented data or causal feature orderings) or when contributions should be adjusted by factors such as trust or risk. We propose Priority-Aware Shapley Value (PASV), which incorporates both hard precedence constraints and soft, contributor-specific priority weights. PASV is applicable to general precedence structures, recovers precedence-only and weight-only Shapley variants as special cases, and is uniquely characterized by natural axioms. We develop an efficient adjacent-swap Metropolis-Hastings sampler for scalable Monte Carlo estimation and analyze limiting regimes induced by extreme priority weights. Experiments on data valuation (MNIST/CIFAR10) and feature attribution (Census Income) demonstrate more structure-faithful allocations and a practical sensitivity analysis via our proposed "priority sweeping".