Challenges of Explaining the Behavior of Black-Box AI Systems

Aleksandre Asatiani; Pekka Malo; Per Rådberg Nagbøl; Esko Penttinen; Tapani Rinta-Kahila; Antti Salovaara

doi:10.17705/2msqe.00037

Challenges of Explaining the Behavior of Black-Box AI Systems

Aleksandre Asatiani, Pekka Malo, Per Rådberg Nagbøl, Esko Penttinen, Tapani Rinta-Kahila, Antti Salovaara

Research output: Journal Article or Conference Article in Journal › Journal article › Research › peer-review

Abstract

There are many examples of problems resulting from inscrutable AI systems, so there is a growing need to be able to explain how such systems produce their outputs. Draw- ing on a case study at the Danish Business Authority, we provide a framework and recommendations for addressing the many challenges of explaining the behavior of black-box AI systems. Our findings will enable organizations to successfully develop and deploy AI systems without causing legal or ethical problems.

Original language	English
Article number	7
Journal	M I S Quarterly Executive
Volume	19
Issue number	4
Pages (from-to)	259-278
ISSN	1540-1960
DOIs	https://doi.org/10.17705/2msqe.00037
Publication status	Published - 1 Dec 2020

Keywords

Artificial intelligence
Explainability
Envelopment

Access to Document

10.17705/2msqe.00037

Challenges of Explaining the Behavior of Black-Box AI SystemsAccepted author manuscript, 2.27 MB

https://aisel.aisnet.org/cgi/viewcontent.cgi?article=1488&context=misqe

Cite this

@article{8a8532e7edb34d0baa913b00e1b254c3,

title = "Challenges of Explaining the Behavior of Black-Box AI Systems",

abstract = "There are many examples of problems resulting from inscrutable AI systems, so there is a growing need to be able to explain how such systems produce their outputs. Draw- ing on a case study at the Danish Business Authority, we provide a framework and recommendations for addressing the many challenges of explaining the behavior of black-box AI systems. Our findings will enable organizations to successfully develop and deploy AI systems without causing legal or ethical problems.",

keywords = "Artificial intelligence, Explainability, Envelopment, Artificial intelligence, Explainability, Envelopment",

author = "Aleksandre Asatiani and Pekka Malo and Nagb{\o}l, {Per R{\aa}dberg} and Esko Penttinen and Tapani Rinta-Kahila and Antti Salovaara",

year = "2020",

month = dec,

day = "1",

doi = "10.17705/2msqe.00037",

language = "English",

volume = "19",

pages = "259--278",

journal = "M I S Quarterly Executive",

issn = "1540-1960",

publisher = "Indiana University Press",

number = "4",

}

TY - JOUR

T1 - Challenges of Explaining the Behavior of Black-Box AI Systems

AU - Asatiani, Aleksandre

AU - Malo, Pekka

AU - Nagbøl, Per Rådberg

AU - Penttinen, Esko

AU - Rinta-Kahila, Tapani

AU - Salovaara, Antti

PY - 2020/12/1

Y1 - 2020/12/1

N2 - There are many examples of problems resulting from inscrutable AI systems, so there is a growing need to be able to explain how such systems produce their outputs. Draw- ing on a case study at the Danish Business Authority, we provide a framework and recommendations for addressing the many challenges of explaining the behavior of black-box AI systems. Our findings will enable organizations to successfully develop and deploy AI systems without causing legal or ethical problems.

AB - There are many examples of problems resulting from inscrutable AI systems, so there is a growing need to be able to explain how such systems produce their outputs. Draw- ing on a case study at the Danish Business Authority, we provide a framework and recommendations for addressing the many challenges of explaining the behavior of black-box AI systems. Our findings will enable organizations to successfully develop and deploy AI systems without causing legal or ethical problems.

KW - Artificial intelligence

KW - Explainability

KW - Envelopment

KW - Artificial intelligence

KW - Explainability

KW - Envelopment

UR - https://aisel.aisnet.org/misqe/vol19/iss4/7/

U2 - 10.17705/2msqe.00037

DO - 10.17705/2msqe.00037

M3 - Journal article

SN - 1540-1960

VL - 19

SP - 259

EP - 278

JO - M I S Quarterly Executive

JF - M I S Quarterly Executive

IS - 4

M1 - 7

ER -

Challenges of Explaining the Behavior of Black-Box AI Systems

Abstract

Keywords

Access to Document

Fingerprint

Cite this