Definition
Explainability is a property of an analytical model or Artificial Intelligence system that describes the degree to which people can understand how or why it produced a particular result or recommendation.
Human Explanation
It helps answer the question: Why did the model produce this result?
Why it Matters
An understandable explanation supports appropriate trust, review and accountability, and helps detect errors, Bias and unsuitable assumptions.
Conceptual Boundary
Explainability does not require every internal calculation to be exposed, and an explanation is not proof that a result is correct. It means that the relevant factors, reasoning or behaviour can be communicated at a level appropriate to the person using or reviewing the result.
Practical Perspective
The greater a model's influence on people, rights, safety or important organisational decisions, the stronger the need for meaningful explanation. An explanation must be accurate enough to support scrutiny, not merely sound plausible.