A Unified Framework for Uncertainty-Aware Explainable Artificial Intelligence: A Case Study in Power Quality Disturbance Classification
arXiv:2605.21114v1 Announce Type: new Abstract: Post-hoc explainable AI (XAI) methods typically produce deterministic attribution maps, whereas Bayesian neural networks (BNNs) induce a distribution over explanations. Capturing the variability of this distribution is important for uncertainty-aware decision-making. This paper formalises the \emph{explanation distribution} as the push-forward measure of the BNN posterior through any Lipschitz-continuous attribution operator. It further proposes the uncertainty-aware relevance attribution operator (UA-RAO), a general family of operators that summ
The increasing deployment of AI in critical systems necessitates robust methods for understanding and quantifying the uncertainty in their decision-making processes.
A unified framework for uncertainty-aware XAI allows for more reliable and trustworthy AI applications, crucial for industries where errors have significant consequences.
AI explanations, previously often deterministic, can now formally incorporate uncertainty, leading to more nuanced and defensible decisions from AI systems.
- · AI developers
- · High-stakes industries (e.g., energy, healthcare, defense)
- · Regulatory bodies
- · Users of AI systems
- · Developers of proprietary, 'black-box' AI solutions without XAI capabilities
- · AI systems prone to unquantified uncertainty
Improved reliability and explainability of AI applications in sensitive domains like power grid management.
Increased adoption of AI in industries previously hesitant due to concerns about trustworthiness and opacity.
Potential for new regulatory standards requiring uncertainty quantification in AI explanations for critical infrastructure.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG