Do Not Trust Additive Explanations (2019-03-27T00:00:00.000000Z)

TL;DR

This paper examines the behavior of the most popular instance-level explanations under the presence of interactions, introduces a new method that detects interactions for instance- level explanations, and performs a large scale benchmark to see how frequently additive explanations may be misleading.

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.

Authors

P. Biecek

6 papers

Alicja Gosiewska

1 papers

TL;DR

Abstract

Authors

References35 items

"How do I fool you?": Manipulating User Trust via Misleading Black Box Explanations

Interpretable Machine Learning

Black-Box vs. White-Box: Understanding Their Advantages and Weaknesses From a Practical Point of View

pyCeterisParibus: explaining Machine Learning models with Ceteris Paribus Profiles in Python

On the (In)fidelity and Sensitivity for Explanations.

Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead

On The Stability of Interpretable Models

Understanding Convolutional Neural Networks for Text Classification

iml: An R package for Interpretable Machine Learning

DALEX: explainers for complex predictive models

On the Robustness of Interpretability Methods

Anchors: High-Precision Model-Agnostic Explanations

Explanations of model predictions with live and breakDown packages

Enslaving the Algorithm: From a “Right to an Explanation” to a “Right to Better Decisions”?

Consistent Individualized Feature Attribution for Tree Ensembles

Counterfactual Explanations Without Opening the Black Box: Automated Decisions and the GDPR

Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy

OpenML Benchmarking Suites and the OpenML100

A causal framework for explaining the predictions of black-box sequence-to-sequence models

A Unified Approach to Interpreting Model Predictions

Learning Important Features Through Propagating Activation Differences

Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization

Algorithmic Transparency via Quantitative Input Influence: Theory and Experiments with Learning Systems

“Why Should I Trust You?”: Explaining the Predictions of Any Classifier

Adverse Events in Robotic Surgery: A Retrospective Study of 14 Years of FDA Data

On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance Propagation

OpenML: networked science in machine learning

Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps

I. INTRODUCTION

Understanding Waterfall Plots

LIME-based Explanations With Interpretable Inputs Based on Ceteris Paribus Profiles

How bad is Sacramento’s air, exactly? Google results appear at odds with reality, some say

When a Computer Program Keeps You in Jail

Journal of the advanced practitioner in oncology

An Applied Perspective on Fairness, Accountability, Transparency, and Explainable AI An Introduction to Machine Learning Interpretability

Field of Study

Journal Information

Name