Problem
Research questions and friction points this paper is trying to address.
Post-deployment drift
Explainable AI (XAI)
local attribution methods
global distribution shifts
drift-induced disparate impact
FMMO: Detecting the Divergence Between Local Attribution and Global Drift
🤖 AI Summary
论文提出FMMO框架，通过结合全局代理模型和模型使用度量，解决局部解释方法在检测模型漂移时的局限性，提高对不公平影响的识别能力。
📝 Abstract
Post-deployment drift poses a critical risk to algorithmic accountability, particularly when ground truth labels are delayed and performance degradation becomes a"silent failure". While Explainable AI (XAI) is often relied upon to audit these shifts, we demonstrate that popular local attribution methods (e.g., TreeSHAP) can exhibit misleading stability even as model reliability collapses. In this paper, we propose a Framework for Model Monitoring and Observability (FMMO) designed to expose the divergence between local explanation stability and global distribution shifts. Using benchmark, synthetic, and real-world datasets, we show that local XAI methods fail to flag drift-induced disparate impact, specifically where False Positive Rates spike for protected groups while feature attributions remain unchanged. By integrating global surrogate models with model utilization measurements, FMMO mitigates this fairness blind spot, ensuring that stakeholders can detect discriminatory deterioration that standard local XAI tools overlook.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Framework for Model Monitoring and Observability (FMMO)
local attribution methods
global distribution shifts
fairness blind spot
drift-induced disparate impact
🔎 Similar Papers
No similar papers found.
