Evaluating Model Explanations without Ground Truth
This addresses the challenge of selecting reliable explanations in machine learning for practitioners, though it is incremental as it builds on existing evaluation principles.
The paper tackles the problem of evaluating model explanations without relying on ground-truth data, proposing a ground-truth agnostic framework (AXE) that provides an independent measure of explanation quality and can detect explanation fairwashing.
There can be many competing and contradictory explanations for a single model prediction, making it difficult to select which one to use. Current explanation evaluation frameworks measure quality by comparing against ideal "ground-truth" explanations, or by verifying model sensitivity to important inputs. We outline the limitations of these approaches, and propose three desirable principles to ground the future development of explanation evaluation strategies for local feature importance explanations. We propose a ground-truth Agnostic eXplanation Evaluation framework (AXE) for evaluating and comparing model explanations that satisfies these principles. Unlike prior approaches, AXE does not require access to ideal ground-truth explanations for comparison, or rely on model sensitivity - providing an independent measure of explanation quality. We verify AXE by comparing with baselines, and show how it can be used to detect explanation fairwashing. Our code is available at https://github.com/KaiRawal/Evaluating-Model-Explanations-without-Ground-Truth.