A Bayesian multilevel analysis of belief alignment effect predicting human moral intuitions of artificial intelligence judgments

Yuxin Liu; Adam Moore

A Bayesian multilevel analysis of belief alignment effect predicting human moral intuitions of artificial intelligence judgments

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract / Description of output

Despite substantial progress in artificial intelligence (AI) little is known about people’s moral intuitions towards AI systems. Given that politico-moral intuitions often influence judgements in non-rational ways, we investigated participants’ willingness to act on verdicts provided by an expert AI system, trust in AI, and perceived fairness of AI as a function of the AI system’s (dis)agreement with their pre-existing politico-moral beliefs across various morally contentious issues. Results show belief alignment triggered a willingness to act on AI verdicts but did not increase trust or fairness perception of the AI. This result was unaffected by general AI attitudes. Our findings suggest a disassociation between acceptance of AI recommendations and judgements of trust/fairness of the AI, and that such acceptance is partly driven by alignment with pre-existing intuitions.

Original language	English
Title of host publication	Proceedings of the 44th Annual Conference of the Cognitive Science Society
Editors	Jennifer Culbertson, Andrew Perfors, Hugh Rabagliati, Veronica Ramenzoni
Publisher	eScholarship University of California
Pages	2116-2125
Publication status	E-pub ahead of print - 17 Jun 2022
Event	44th Annual Meeting of the Cognitive Science Society - Toronto, Canada Duration: 27 Jul 2022 → 30 Jul 2022 Conference number: 44 https://cognitivesciencesociety.org/cogsci-2022/

Publication series

Name	Proceedings of the Annual Conference of the Cognitive Science Society
Publisher	Cognitive Science Society
Volume	44
ISSN (Electronic)	1069-7977

Conference

Conference	44th Annual Meeting of the Cognitive Science Society
Abbreviated title	CogSci 2022
Country/Territory	Canada
City	Toronto
Period	27/07/22 → 30/07/22
Internet address	https://cognitivesciencesociety.org/cogsci-2022/

Keywords / Materials (for Non-textual outputs)

artificial intelligence
human-AI interaction
moral intuitions
belief alignment
political partisanship

Access to Document

LiuMoore2022ABaysianMultilevelAnalysisFinal published version, 486 KBLicence: Creative Commons: Attribution (CC-BY)

https://escholarship.org/uc/item/3v79704hLicence: Creative Commons: Attribution (CC-BY)

1 Book

Proceedings of the 44th Annual Conference of the Cognitive Science Society
Culbertson, J. (ed.), Perfors, A. (ed.), Rabagliati, H. (ed.) & Ramenzoni, V. (ed.), Jul 2022, (E-pub ahead of print) eScholarship University of California. (Proceedings of the Annual Conference of the Cognitive Science Society; vol. 44)
Research output: Book/Report › Book

Open Access

Cite this

Liu, Y., & Moore, A. (2022). A Bayesian multilevel analysis of belief alignment effect predicting human moral intuitions of artificial intelligence judgments. In J. Culbertson, A. Perfors, H. Rabagliati, & V. Ramenzoni (Eds.), Proceedings of the 44th Annual Conference of the Cognitive Science Society (pp. 2116-2125). (Proceedings of the Annual Conference of the Cognitive Science Society; Vol. 44). eScholarship University of California. Advance online publication. https://escholarship.org/uc/item/3v79704h

Liu, Yuxin ; Moore, Adam. / A Bayesian multilevel analysis of belief alignment effect predicting human moral intuitions of artificial intelligence judgments. Proceedings of the 44th Annual Conference of the Cognitive Science Society. editor / Jennifer Culbertson ; Andrew Perfors ; Hugh Rabagliati ; Veronica Ramenzoni. eScholarship University of California, 2022. pp. 2116-2125 (Proceedings of the Annual Conference of the Cognitive Science Society).

@inproceedings{a3bf5682c16e44df9b8a8d86dd1b3ee3,

title = "A Bayesian multilevel analysis of belief alignment effect predicting human moral intuitions of artificial intelligence judgments",

abstract = "Despite substantial progress in artificial intelligence (AI) little is known about people{\textquoteright}s moral intuitions towards AI systems. Given that politico-moral intuitions often influence judgements in non-rational ways, we investigated participants{\textquoteright} willingness to act on verdicts provided by an expert AI system, trust in AI, and perceived fairness of AI as a function of the AI system{\textquoteright}s (dis)agreement with their pre-existing politico-moral beliefs across various morally contentious issues. Results show belief alignment triggered a willingness to act on AI verdicts but did not increase trust or fairness perception of the AI. This result was unaffected by general AI attitudes. Our findings suggest a disassociation between acceptance of AI recommendations and judgements of trust/fairness of the AI, and that such acceptance is partly driven by alignment with pre-existing intuitions.",

keywords = "artificial intelligence, human-AI interaction, moral intuitions, belief alignment, political partisanship",

author = "Yuxin Liu and Adam Moore",

year = "2022",

month = jun,

day = "17",

language = "English",

series = "Proceedings of the Annual Conference of the Cognitive Science Society",

publisher = "eScholarship University of California",

pages = "2116--2125",

editor = "Jennifer Culbertson and Andrew Perfors and Hugh Rabagliati and Veronica Ramenzoni",

booktitle = "Proceedings of the 44th Annual Conference of the Cognitive Science Society",

address = "United States",

note = "44th Annual Meeting of the Cognitive Science Society, CogSci 2022 ; Conference date: 27-07-2022 Through 30-07-2022",

url = "https://cognitivesciencesociety.org/cogsci-2022/",

}

Liu, Y & Moore, A 2022, A Bayesian multilevel analysis of belief alignment effect predicting human moral intuitions of artificial intelligence judgments. in J Culbertson, A Perfors, H Rabagliati & V Ramenzoni (eds), Proceedings of the 44th Annual Conference of the Cognitive Science Society. Proceedings of the Annual Conference of the Cognitive Science Society, vol. 44, eScholarship University of California, pp. 2116-2125, 44th Annual Meeting of the Cognitive Science Society, Toronto, Canada, 27/07/22. <https://escholarship.org/uc/item/3v79704h>

A Bayesian multilevel analysis of belief alignment effect predicting human moral intuitions of artificial intelligence judgments. / Liu, Yuxin; Moore, Adam.
Proceedings of the 44th Annual Conference of the Cognitive Science Society. ed. / Jennifer Culbertson; Andrew Perfors; Hugh Rabagliati; Veronica Ramenzoni. eScholarship University of California, 2022. p. 2116-2125 (Proceedings of the Annual Conference of the Cognitive Science Society; Vol. 44).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - A Bayesian multilevel analysis of belief alignment effect predicting human moral intuitions of artificial intelligence judgments

AU - Liu, Yuxin

AU - Moore, Adam

N1 - Conference code: 44

PY - 2022/6/17

Y1 - 2022/6/17

N2 - Despite substantial progress in artificial intelligence (AI) little is known about people’s moral intuitions towards AI systems. Given that politico-moral intuitions often influence judgements in non-rational ways, we investigated participants’ willingness to act on verdicts provided by an expert AI system, trust in AI, and perceived fairness of AI as a function of the AI system’s (dis)agreement with their pre-existing politico-moral beliefs across various morally contentious issues. Results show belief alignment triggered a willingness to act on AI verdicts but did not increase trust or fairness perception of the AI. This result was unaffected by general AI attitudes. Our findings suggest a disassociation between acceptance of AI recommendations and judgements of trust/fairness of the AI, and that such acceptance is partly driven by alignment with pre-existing intuitions.

AB - Despite substantial progress in artificial intelligence (AI) little is known about people’s moral intuitions towards AI systems. Given that politico-moral intuitions often influence judgements in non-rational ways, we investigated participants’ willingness to act on verdicts provided by an expert AI system, trust in AI, and perceived fairness of AI as a function of the AI system’s (dis)agreement with their pre-existing politico-moral beliefs across various morally contentious issues. Results show belief alignment triggered a willingness to act on AI verdicts but did not increase trust or fairness perception of the AI. This result was unaffected by general AI attitudes. Our findings suggest a disassociation between acceptance of AI recommendations and judgements of trust/fairness of the AI, and that such acceptance is partly driven by alignment with pre-existing intuitions.

KW - artificial intelligence

KW - human-AI interaction

KW - moral intuitions

KW - belief alignment

KW - political partisanship

M3 - Conference contribution

T3 - Proceedings of the Annual Conference of the Cognitive Science Society

SP - 2116

EP - 2125

BT - Proceedings of the 44th Annual Conference of the Cognitive Science Society

A2 - Culbertson, Jennifer

A2 - Perfors, Andrew

A2 - Rabagliati, Hugh

A2 - Ramenzoni, Veronica

PB - eScholarship University of California

T2 - 44th Annual Meeting of the Cognitive Science Society

Y2 - 27 July 2022 through 30 July 2022

ER -

Liu Y, Moore A. A Bayesian multilevel analysis of belief alignment effect predicting human moral intuitions of artificial intelligence judgments. In Culbertson J, Perfors A, Rabagliati H, Ramenzoni V, editors, Proceedings of the 44th Annual Conference of the Cognitive Science Society. eScholarship University of California. 2022. p. 2116-2125. (Proceedings of the Annual Conference of the Cognitive Science Society). Epub 2022 Jun 17.

A Bayesian multilevel analysis of belief alignment effect predicting human moral intuitions of artificial intelligence judgments

Abstract / Description of output

Publication series

Conference

Keywords / Materials (for Non-textual outputs)

Access to Document

Fingerprint

Research output

Proceedings of the 44th Annual Conference of the Cognitive Science Society

Cite this