Skip to content
BENIBIX Learn

Explainable AI (XAI)

Also known as XAI, AI Explainability, Explainability

Ways to show people why an AI made a choice.

Think of it like

It is like a teacher who shows you how they marked your test.

Example

Lucy's loan request is turned down by an app. The app shows that her high debt and two missed payments weighed most, so she knows what to fix before trying again.

Why it matters

People trust what they understand. They can only push back on a choice if they know the reason.

Definition

Explainable AI is a set of methods that show people why an AI made a choice. A model may turn down a loan, and the person asks why. An explainable system can point to the inputs that weighed most, such as income or missed payments. Common tools score how much each input pushed the answer up or down. Some laws give people a right to a real reason when a machine's choice affects them.
The answer, and the three reasons that weighed most.

Common mistake

Taking the reason as the whole truth. Most tools give a short summary of a big model, and the summary can mislead.

Key takeaways

  1. 1
    Why, not just what. Show the reason.
  2. 2
    Inputs that mattered. Ranked.
  3. 3
    A summary. Not the whole truth.

Compare