<ul><li>Machine learning explainability focuses on making black-box models more transparent by identifying important input features for predictions.</li><li>Recent research suggests creating explanations using semantic concepts to enhance interpretability, but often overlooks the communication context.</li><li>This study introduces listener-adaptive explanations, which consider the user's understanding to maximize communicative utility.</li><li>The proposed method, based on pragmatic reasoning, showed improved alignment with listener preferences in image classification tasks and increased communicative utility in user studies.</li></ul>

Aligning Explanations with Human Communication

Discover more