ExplainabilityTransparencyAccountabilitystat.ML
Explainable AI by BAPC -- Before and After correction Parameter Comparison
Florian Sobieczky, Manuela Geiß
March 12, 2021
arXiv.org
3 citations
Key Takeaways
-
Explainability through parameter comparison: The method explains AI predictions by showing how a complex model changes the parameters of a simpler, interpretable base model (linear regression), making it easier to understand what the AI is doing differently compared to a straightforward approach.
-
Balancing accuracy and interpretability: The research establishes specific criteria that balance three competing goals—maintaining surrogate accuracy, preserving the AI model's performance, and ensuring the explanation faithfully represents the model's behavior—rather than treating them as separate concerns.
-
Context-dependent explanations: The approach identifies that explanations work best within specific neighborhoods around each prediction, with an "ideal size" that depends on the noise level in the data, suggesting that one-size-fits-all explanations may be less reliable than localized ones.
-
Practical framework for auditing: By decomposing AI predictions into "before and after" parameter changes, this method provides a concrete way for non-technical stakeholders to audit whether an AI system's decisions deviate from simpler, more understandable baseline models in justifiable ways.
Abstract
A local surrogate for an AI-model correcting a simpler 'base' model is introduced representing an analytical method to yield explanations of AI-predictions. The approach is studied here in the context of the base model being linear regression. The AI-model approximates the residual error of the linear model and the explanations are formulated in terms of the change of the interpretable base model's parameters. Criteria are formulated for the precise relation between lost accuracy of the surrogate, the accuracy of the AI-model, and the surrogate fidelity. It is shown that, assuming a certain maximal amount of noise in the observed data, these criteria induce neighborhoods of the instances to be explained which have an ideal size in terms of maximal accuracy and fidelity.