Identifying systematic errors or unfair patterns in AI model predictions across different groups or categories.