Abstract:
The
Right to Explanation is an important regulatory principle that allows individuals to request actionable explanations for algorithmic decisions. However, several technical challenges arise when providing such actionable explanations in practice. For instance,
models are periodically retrained to handle dataset shifts, and this may in turn invalidate some of the previously prescribed explanations thus rendering them unactionable. However, it is unclear if and when such invalidations occur, and what factors determine
explanation stability i.e., if an explanation remains unchanged amidst model retraining due to dataset shifts. In this talk, I will discuss how my work from my summer internship aims to address this problem by providing a theoretical and empirical characterization
of the factors influencing explanation stability. In particular, we conduct a theoretical analysis to show that model curvature, weight decay parameters, and the magnitude of the dataset shift are key factors that determine the extent of explanation (in)stability.
Our experiments on real-world datasets not only validate these theoretical results, but also demonstrate that the aforementioned factors dramatically impact the stability of explanations produced by various state-of-the-art methods.
![[Computer Systems Lab]](http://www.cs.wisc.edu/pics/csl_logo.gif){kind=logo}