Problems with Shapley-value-based explanations as feature importance measures

Authors: I. Elizabeth Kumar, Suresh Venkatasubramanian, Carlos Scheidegger, Sorelle Friedler

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical In this paper, we demonstrate that Shapley-value-based explanations for feature importance fail to serve their desired purpose in general. We make this argument in two parts. Firstly, we show that applying the Shapley value to the problem of feature importance introduces mathematically formalizable properties which may not align with what we would expect from an explanation. Secondly, taking a human-centric perspective, we evaluate Shapley-value-based explanations through established frameworks of what people expect from explanations, and find them wanting.
Researcher Affiliation Academia 1School of Computing, University of Utah, Salt Lake City, UT, USA 2Department of Computer Science, University of Arizona, Tucson, AZ, USA 3Department of Computer Science, Haverford College, Haverford, PA, USA.
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide concrete access to source code for a methodology described by the authors, as it is a theoretical critique rather than a new system implementation.
Open Datasets No The paper is a theoretical analysis and does not involve training models on datasets, thus it does not provide concrete access information for a publicly available or open dataset for its own work.
Dataset Splits No The paper is theoretical and does not conduct experiments, so it does not provide specific dataset split information.
Hardware Specification No The paper is theoretical and does not describe specific hardware details used for running experiments.
Software Dependencies No The paper is theoretical and does not specify software dependencies with version numbers.
Experiment Setup No The paper is theoretical and does not describe specific experimental setup details or hyperparameters.