Explainable artificial intelligence (XAI), the goodness criteria and the grasp-ability test

Explainability needs a measurable standard: The paper proposes the "grasp-ability test" as a concrete way to evaluate whether AI explanations actually help users understand how algorithms work, rather than assuming all explanations are equally useful.

Not all explanations are created equal: The research emphasizes that different explanations vary in their meaningfulness to users, suggesting that organizations should actively compare and select explanations based on how well people can actually comprehend them.

Understanding matters for accountability: By focusing on whether users can "grasp" how algorithmic decisions are made, the paper highlights that explainability serves a practical purpose—enabling people to understand and potentially challenge automated decisions that affect them.

Goodness criteria should be user-centered: The framework suggests that the quality of an AI explanation should be judged by whether real users find it meaningful and understandable, not just by technical metrics or what developers think is clear.