Explainability is the property of an AI system to present the factors behind a result in a way that people can follow. For simple models such as a decision tree, this follows from the logic itself. For complex models such as neural networks, post-hoc methods like SHAP or LIME show which features had the strongest influence on an individual output.
Explainability is a precondition for employees to check a result and for affected persons to challenge it. The AI Act requires that deployers can interpret the output of high-risk AI systems (Art. 13) and gives persons affected by certain decisions a right to an explanation (Art. 86). For automated individual decisions, the GDPR requires that data subjects receive meaningful information on how the decision logic works. More under T02 Lack of transparency and explainability.