The Promise and Peril of Human Evaluation for Model Interpretability

Bernease Herman

Transparency, user trust, and human comprehension are popular ethical motivations for interpretable machine learning. In support of these goals, researchers evaluate model explanat…