Contextual Importance and Utility: A Theoretical Foundation

Kary Främling*

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionScientificpeer-review

Abstract

This paper provides new theory to support to the eXplainable AI (XAI) method Contextual Importance and Utility (CIU). CIU arithmetic is based on the concepts of Multi-Attribute Utility Theory, which gives CIU a solid theoretical foundation. The novel concept of contextual influence is also defined, which makes it possible to compare CIU directly with so-called additive feature attribution (AFA) methods for model-agnostic outcome explanation. One key takeaway is that the ‘influence’ concept used by AFA methods is inadequate for outcome explanation purposes even for simple models to explain. Experiments with simple models show that explanations using contextual importance (CI) and contextual utility (CU) produce explanations where influence-based methods fail. It is also shown that CI and CU guarantees explanation faithfulness towards the explained model.

Original languageEnglish
Title of host publicationAI 2021
Subtitle of host publicationAdvances in Artificial Intelligence - 34th Australasian Joint Conference, AI 2021, Proceedings
EditorsGuodong Long, Xinghuo Yu, Sen Wang
Pages117-128
Number of pages12
DOIs
Publication statusPublished - 2022
MoE publication typeA4 Article in a conference publication
EventAustralasian Joint Conference on Artificial Intelligence - Virtual, Online, Australia
Duration: 2 Feb 20224 Feb 2022
Conference number: 34

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PublisherSpringer
Volume13151 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

ConferenceAustralasian Joint Conference on Artificial Intelligence
Abbreviated titleAI
Country/TerritoryAustralia
CityVirtual, Online
Period02/02/202204/02/2022

Keywords

  • Contextual importance and utility
  • Decision theory
  • Explainable AI
  • Multi-attribute utility theory

Fingerprint

Dive into the research topics of 'Contextual Importance and Utility: A Theoretical Foundation'. Together they form a unique fingerprint.

Cite this