Interpretability
The degree to which humans can understand and trace an AI system's reasoning process and the factors that led to its outputs.
Interpretability encompasses techniques and design choices that make AI decision-making transparent and auditable. In scientific contexts, interpretability is especially critical because researchers must be able to evaluate, trace, and build upon AI-generated reasoning. A prediction that cannot be interrogated is scientifically worthless regardless of its accuracy.
Also known as
explainability, AI transparency, model interpretability