Factors Supporting Explanation Learning

Background: The process by which users learn the decision pattern of an AI system based on its explanations is a key component of human understanding.

Question / Future Work: Identify the specific factors beyond functional correctness that are necessary to support users in learning the AI’s decision pattern, given that even fully correct explanations did not guarantee understanding for all participants. This work should aim to establish a comprehensive set of necessary conditions for explanation-based learning.

Why It Matters: Correctness is a necessary but not sufficient condition for understanding; identifying the other crucial factors (e.g., presentation style, required cognitive effort) is required to move XAI evaluation beyond simply measuring faithfulness.

Evidence: At the same time, correctness alone was not sufficient for understanding, as many participants failed to learn even from fully correct explanations. … Given that correctness alone did not suffice for understanding, identifying what other factors support learning from explanations is equally important.