Explainable Artificial Intelligence (XAI)has received a great deal of
attention recently. Explainability is being presented as a remedy for the
distrust of complex and opaque models. Model agnostic methods such as LIME,
SHAP, or Break Down promise instance-level interpretability for any complex
machine learning model. But how faithful are these additive explanations? Can
we rely on additive explanations for non-additive models?
In this paper, we (1) examine the behavior of the most popular instance-level
explanations under the presence of interactions, (2) introduce a new method
that detects interactions for instance-level explanations, (3) perform a large
scale benchmark to see how frequently additive explanations may be misleading.
%0 Journal Article
%1 gosiewska2019trust
%A Gosiewska, Alicja
%A Biecek, Przemyslaw
%D 2019
%K xai
%T Do Not Trust Additive Explanations
%U http://arxiv.org/abs/1903.11420
%X Explainable Artificial Intelligence (XAI)has received a great deal of
attention recently. Explainability is being presented as a remedy for the
distrust of complex and opaque models. Model agnostic methods such as LIME,
SHAP, or Break Down promise instance-level interpretability for any complex
machine learning model. But how faithful are these additive explanations? Can
we rely on additive explanations for non-additive models?
In this paper, we (1) examine the behavior of the most popular instance-level
explanations under the presence of interactions, (2) introduce a new method
that detects interactions for instance-level explanations, (3) perform a large
scale benchmark to see how frequently additive explanations may be misleading.
@article{gosiewska2019trust,
abstract = {Explainable Artificial Intelligence (XAI)has received a great deal of
attention recently. Explainability is being presented as a remedy for the
distrust of complex and opaque models. Model agnostic methods such as LIME,
SHAP, or Break Down promise instance-level interpretability for any complex
machine learning model. But how faithful are these additive explanations? Can
we rely on additive explanations for non-additive models?
In this paper, we (1) examine the behavior of the most popular instance-level
explanations under the presence of interactions, (2) introduce a new method
that detects interactions for instance-level explanations, (3) perform a large
scale benchmark to see how frequently additive explanations may be misleading.},
added-at = {2021-12-30T00:46:37.000+0100},
author = {Gosiewska, Alicja and Biecek, Przemyslaw},
biburl = {https://www.bibsonomy.org/bibtex/2dabf149e1b31d5c6a7db5bd58b86098a/alexv},
description = {[1903.11420] Do Not Trust Additive Explanations},
interhash = {ae9d2e16584fcbe7f2d90f5e83a22fd2},
intrahash = {dabf149e1b31d5c6a7db5bd58b86098a},
keywords = {xai},
note = {cite arxiv:1903.11420Comment: 15 pages},
timestamp = {2021-12-30T00:46:37.000+0100},
title = {Do Not Trust Additive Explanations},
url = {http://arxiv.org/abs/1903.11420},
year = 2019
}