Explainable Predictive Maintenance

One-size-fits-all explanations don't work: Current XAI research focuses mainly on building user trust or debugging models, but fails to recognize that different users (data scientists, domain experts, end-users) and different industries need tailored explanations suited to their specific contexts and needs.

Predictive maintenance is an overlooked application area: Despite explainability being critical for industrial applications—particularly in predictive maintenance under Industry 4.0 and 5.0—this domain remains "relatively under-explored," representing a significant gap between what XAI research currently offers and what industry actually needs.

The "black box" problem requires context-specific solutions: While XAI helps make complex machine learning decisions more understandable, the paper emphasizes that generic explainability techniques may not adequately address the diverse purposes explanations serve across different industrial settings and user groups.

Explanations serve multiple purposes beyond trust: The paper identifies that explanations are needed not just to build user confidence, but also for debugging models and meeting domain-specific requirements—suggesting that responsible AI implementation requires understanding why an explanation is needed before choosing how to provide it.