AI systems designed to provide human-understandable justifications for their decisions and actions, enhancing transparency and trust.
Techniques and frameworks that generate clear, context-appropriate explanations - feature attributions, rule extractions, counterfactual scenarios - tailored to stakeholder needs (end users, regulators, developers). XAI governance includes standardizing explanation formats, validating explanation fidelity, and training users to interpret them properly.
A credit-card company uses XAI by integrating LIME explanations into its fraud alerts: when the AI flags a transaction as suspicious, it shows the user that “Unusual merchant location” and “High transaction amount” were the top contributors, enabling faster verification and reducing false-positive escalations by 30%.
