Missingness Bias Calibration in Feature Attribution Explanations
This work provides a more efficient way to improve the reliability of feature attribution explanations for researchers and practitioners using AI models in medical domains, offering an incremental improvement over existing solutions.
This paper addresses missingness bias in feature attribution explanations, which causes unreliable importance scores when models are probed with ablated inputs. The authors introduce MCal, a lightweight post-hoc method that fine-tunes a linear head on the frozen base model's outputs to correct this bias, achieving competitive or superior performance compared to prior heavyweight approaches across various medical benchmarks.
Popular explanation methods often produce unreliable feature importance scores due to missingness bias, a systematic distortion that arises when models are probed with ablated, out-of-distribution inputs. Existing solutions treat this as a deep representational flaw that requires expensive retraining or architectural modifications. In this work, we challenge this assumption and show that missingness bias can be effectively treated as a superficial artifact of the model's output space. We introduce MCal, a lightweight post-hoc method that corrects this bias by fine-tuning a simple linear head on the outputs of a frozen base model. Surprisingly, we find this simple correction consistently reduces missingness bias and is competitive with, or even outperforms, prior heavyweight approaches across diverse medical benchmarks spanning vision, language, and tabular domains.