Normative Monitoring Using Bayesian Networks: Defining a Threshold for Conflict Detection

Annet Onnes; Silja Renooij; Mehdi Dastani

doi:10.1007/978-3-031-45608-4_12

Normative Monitoring Using Bayesian Networks: Defining a Threshold for Conflict Detection

Annet Onnes^*, Silja Renooij, Mehdi Dastani

^*Corresponding author for this work

Sub Intelligent Systems

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

Abstract

Normative monitoring of black-box AI systems entails detecting whether input-output combinations of AI systems are acceptable in specific contexts. To this end, we build on an existing approach that uses Bayesian networks and a tailored conflict measure called IOconfl. In this paper, we argue that the default fixed threshold associated with this measure is not necessarily suitable for the purpose of normative monitoring. We subsequently study the bounds imposed on the measure by the normative setting and, based upon our analyses, propose a dynamic threshold that depends on the context in which the AI system is applied. Finally, we show the measure and threshold are effective by experimentally evaluating them using an existing Bayesian network.

Original language	English
Title of host publication	Symbolic and Quantitative Approaches to Reasoning with Uncertainty
Subtitle of host publication	17th European Conference, ECSQARU 2023, Arras, France, September 19–22, 2023, Proceedings
Editors	Zied Bouraoui, Srdjan Vesic
Publisher	Springer
Pages	149–159
Number of pages	11
ISBN (Electronic)	978-3-031-45608-4
ISBN (Print)	9783031456077
DOIs	https://doi.org/10.1007/978-3-031-45608-4_12
Publication status	Published - 19 Nov 2023

Publication series

Name	Lecture Notes in Computer Science
Publisher	Springer
Volume	14294
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Bibliographical note

Publisher Copyright:
© 2024, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Funding

This research was supported by the Hybrid Intelligence Centre, a 10-year programme funded by the Dutch Ministry of Education, Culture and Science through the Netherlands Organisation for Scientific Research, https://hybrid-intelligence-centre.nl.

Funders	Funder number
Hybrid Intelligence Centre
Ministerie van onderwijs, cultuur en wetenschap
Nederlandse Organisatie voor Wetenschappelijk Onderzoek

Keywords

Bayesian Networks
Conflict Analysis
Normative Monitoring
Responsible AI

Access to Document

10.1007/978-3-031-45608-4_12

978-3-031-45608-4_12Final published version, 512 KBLicence: Taverne

Cite this

Onnes, A., Renooij, S., & Dastani, M. (2023). Normative Monitoring Using Bayesian Networks: Defining a Threshold for Conflict Detection. In Z. Bouraoui, & S. Vesic (Eds.), Symbolic and Quantitative Approaches to Reasoning with Uncertainty: 17th European Conference, ECSQARU 2023, Arras, France, September 19–22, 2023, Proceedings (pp. 149–159). (Lecture Notes in Computer Science; Vol. 14294). Springer. https://doi.org/10.1007/978-3-031-45608-4_12

Onnes, Annet ; Renooij, Silja ; Dastani, Mehdi. / Normative Monitoring Using Bayesian Networks: Defining a Threshold for Conflict Detection. Symbolic and Quantitative Approaches to Reasoning with Uncertainty: 17th European Conference, ECSQARU 2023, Arras, France, September 19–22, 2023, Proceedings. editor / Zied Bouraoui ; Srdjan Vesic. Springer, 2023. pp. 149–159 (Lecture Notes in Computer Science).

@inproceedings{066a468627644161b0cae0672962f506,

title = "Normative Monitoring Using Bayesian Networks: Defining a Threshold for Conflict Detection",

abstract = "Normative monitoring of black-box AI systems entails detecting whether input-output combinations of AI systems are acceptable in specific contexts. To this end, we build on an existing approach that uses Bayesian networks and a tailored conflict measure called IOconfl. In this paper, we argue that the default fixed threshold associated with this measure is not necessarily suitable for the purpose of normative monitoring. We subsequently study the bounds imposed on the measure by the normative setting and, based upon our analyses, propose a dynamic threshold that depends on the context in which the AI system is applied. Finally, we show the measure and threshold are effective by experimentally evaluating them using an existing Bayesian network.",

keywords = "Bayesian Networks, Conflict Analysis, Normative Monitoring, Responsible AI",

author = "Annet Onnes and Silja Renooij and Mehdi Dastani",

note = "Publisher Copyright: {\textcopyright} 2024, The Author(s), under exclusive license to Springer Nature Switzerland AG.",

year = "2023",

month = nov,

day = "19",

doi = "10.1007/978-3-031-45608-4_12",

language = "English",

isbn = "9783031456077",

series = "Lecture Notes in Computer Science",

publisher = "Springer",

pages = "149–159",

editor = "Zied Bouraoui and Srdjan Vesic",

booktitle = "Symbolic and Quantitative Approaches to Reasoning with Uncertainty",

}

Onnes, A , Renooij, S & Dastani, M 2023, Normative Monitoring Using Bayesian Networks: Defining a Threshold for Conflict Detection. in Z Bouraoui & S Vesic (eds), Symbolic and Quantitative Approaches to Reasoning with Uncertainty: 17th European Conference, ECSQARU 2023, Arras, France, September 19–22, 2023, Proceedings. Lecture Notes in Computer Science, vol. 14294, Springer, pp. 149–159. https://doi.org/10.1007/978-3-031-45608-4_12

Normative Monitoring Using Bayesian Networks: Defining a Threshold for Conflict Detection. / Onnes, Annet ; Renooij, Silja ; Dastani, Mehdi.
Symbolic and Quantitative Approaches to Reasoning with Uncertainty: 17th European Conference, ECSQARU 2023, Arras, France, September 19–22, 2023, Proceedings. ed. / Zied Bouraoui; Srdjan Vesic. Springer, 2023. p. 149–159 (Lecture Notes in Computer Science; Vol. 14294).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - Normative Monitoring Using Bayesian Networks: Defining a Threshold for Conflict Detection

AU - Onnes, Annet

AU - Renooij, Silja

AU - Dastani, Mehdi

PY - 2023/11/19

Y1 - 2023/11/19

N2 - Normative monitoring of black-box AI systems entails detecting whether input-output combinations of AI systems are acceptable in specific contexts. To this end, we build on an existing approach that uses Bayesian networks and a tailored conflict measure called IOconfl. In this paper, we argue that the default fixed threshold associated with this measure is not necessarily suitable for the purpose of normative monitoring. We subsequently study the bounds imposed on the measure by the normative setting and, based upon our analyses, propose a dynamic threshold that depends on the context in which the AI system is applied. Finally, we show the measure and threshold are effective by experimentally evaluating them using an existing Bayesian network.

AB - Normative monitoring of black-box AI systems entails detecting whether input-output combinations of AI systems are acceptable in specific contexts. To this end, we build on an existing approach that uses Bayesian networks and a tailored conflict measure called IOconfl. In this paper, we argue that the default fixed threshold associated with this measure is not necessarily suitable for the purpose of normative monitoring. We subsequently study the bounds imposed on the measure by the normative setting and, based upon our analyses, propose a dynamic threshold that depends on the context in which the AI system is applied. Finally, we show the measure and threshold are effective by experimentally evaluating them using an existing Bayesian network.

KW - Bayesian Networks

KW - Conflict Analysis

KW - Normative Monitoring

KW - Responsible AI

UR - http://www.scopus.com/inward/record.url?scp=85177883317&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-45608-4_12

DO - 10.1007/978-3-031-45608-4_12

M3 - Conference contribution

SN - 9783031456077

T3 - Lecture Notes in Computer Science

SP - 149

EP - 159

BT - Symbolic and Quantitative Approaches to Reasoning with Uncertainty

A2 - Bouraoui, Zied

A2 - Vesic, Srdjan

PB - Springer

ER -

Onnes A , Renooij S , Dastani M. Normative Monitoring Using Bayesian Networks: Defining a Threshold for Conflict Detection. In Bouraoui Z, Vesic S, editors, Symbolic and Quantitative Approaches to Reasoning with Uncertainty: 17th European Conference, ECSQARU 2023, Arras, France, September 19–22, 2023, Proceedings. Springer. 2023. p. 149–159. (Lecture Notes in Computer Science). doi: 10.1007/978-3-031-45608-4_12

Normative Monitoring Using Bayesian Networks: Defining a Threshold for Conflict Detection

Abstract

Publication series

Bibliographical note

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this