Rigorous Interpretation Is a Form of Evaluation
For the machine learning community, this paper provides a conceptual framework for using interpretability as a rigorous evaluation tool, though it is primarily a position piece without empirical results.
The paper argues that interpretability can serve as a richer form of model evaluation beyond surface-level metrics, identifying root causes of failures, detecting faulty mechanisms, and predicting issues. It proposes that interpretability must meet scientific standards of falsifiability, reproducibility, and predictiveness.
Current machine learning models are evaluated through behavioral snapshots, with benchmark accuracies, win rates and outcome-based metrics. Model explanations and evaluations, however, are fundamentally intertwined: understanding why a model produces a behavior can be as important as measuring what it produces. If we trusted interpretability, we argue that it can serve not merely as diagnostics but as a richer and more principled form of model evaluation beyond surface-level performance metrics. We explore three ways interpretability can function evaluatively: (1) fixing problems by identifying the root causes of unwanted behavior, (2) detecting subtly faulty mechanisms that invalidate model outputs, and (3) predicting potential issues before they arise by fully understanding the model's weaknesses. To fulfill its evaluative potential, we argue that interpretability methods must generate claims that are falsifiable, reproducible, and predictive -- that is, interpretability must meet scientific standards.