Position: We can measure XAI explanations better with templates

Jonathan Dodge; Margaret Burnett

Position: We can measure XAI explanations better with templates

Jonathan Dodge
, Margaret Burnett

College of Information Sciences and Technology

Research output: Contribution to journal › Conference article › peer-review

3 Scopus citations

Abstract

This paper argues that the Explainable AI (XAI) research community needs to think harder about how to compare, measure, and describe the quality of XAI explanations. We conclude that one (or a few) explanations can be reasonably assessed with methods of the “Explanation Satisfaction” type, but that scaling up our ability to evaluate explanations requires more development of “Explanation Goodness” methods.

Original language	English (US)
Journal	CEUR Workshop Proceedings
Volume	2582
State	Published - 2020
Event	2020 Workshop on Explainable Smart Systems for Algorithmic Transparency in Emerging Technologies, ExSS-ATEC 2020 - Cagliari, Italy Duration: Mar 17 2020 → …

All Science Journal Classification (ASJC) codes

General Computer Science

Cite this

@article{5966f658bf13456294c579371af1e379,

title = "Position: We can measure XAI explanations better with templates",

abstract = "This paper argues that the Explainable AI (XAI) research community needs to think harder about how to compare, measure, and describe the quality of XAI explanations. We conclude that one (or a few) explanations can be reasonably assessed with methods of the “Explanation Satisfaction” type, but that scaling up our ability to evaluate explanations requires more development of “Explanation Goodness” methods.",

author = "Jonathan Dodge and Margaret Burnett",

note = "Funding Information: This work was supported by DARPA \#N66001-17-2-4030. We would like to acknowledge all our co-authors on the work cited here, with specific highlights to Andrew Anderson, Alan Fern, Q. Vera Liao, Yunfeng Zhang, Rachel Bellamy, and Casey Dugan—research work does not happen in a vacuum, nor do ideas. Publisher Copyright: Copyright {\textcopyright} 2020 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).; 2020 Workshop on Explainable Smart Systems for Algorithmic Transparency in Emerging Technologies, ExSS-ATEC 2020 ; Conference date: 17-03-2020",

year = "2020",

language = "English (US)",

volume = "2582",

journal = "CEUR Workshop Proceedings",

issn = "1613-0073",

publisher = "CEUR-WS",

}

TY - JOUR

T1 - Position

T2 - 2020 Workshop on Explainable Smart Systems for Algorithmic Transparency in Emerging Technologies, ExSS-ATEC 2020

AU - Dodge, Jonathan

AU - Burnett, Margaret

N1 - Funding Information: This work was supported by DARPA #N66001-17-2-4030. We would like to acknowledge all our co-authors on the work cited here, with specific highlights to Andrew Anderson, Alan Fern, Q. Vera Liao, Yunfeng Zhang, Rachel Bellamy, and Casey Dugan—research work does not happen in a vacuum, nor do ideas. Publisher Copyright: Copyright © 2020 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).

PY - 2020

Y1 - 2020

N2 - This paper argues that the Explainable AI (XAI) research community needs to think harder about how to compare, measure, and describe the quality of XAI explanations. We conclude that one (or a few) explanations can be reasonably assessed with methods of the “Explanation Satisfaction” type, but that scaling up our ability to evaluate explanations requires more development of “Explanation Goodness” methods.

AB - This paper argues that the Explainable AI (XAI) research community needs to think harder about how to compare, measure, and describe the quality of XAI explanations. We conclude that one (or a few) explanations can be reasonably assessed with methods of the “Explanation Satisfaction” type, but that scaling up our ability to evaluate explanations requires more development of “Explanation Goodness” methods.

UR - https://www.scopus.com/pages/publications/85082951728

UR - https://www.scopus.com/pages/publications/85082951728#tab=citedBy

M3 - Conference article

AN - SCOPUS:85082951728

SN - 1613-0073

VL - 2582

JO - CEUR Workshop Proceedings

JF - CEUR Workshop Proceedings

Y2 - 17 March 2020

ER -

Position: We can measure XAI explanations better with templates

Abstract

All Science Journal Classification (ASJC) codes

Other files and links

Fingerprint

Cite this