John praised Mary because he ? Implicit Causality Bias and Its Interaction with Explicit Cues in LMs

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

Standard

John praised Mary because he ? Implicit Causality Bias and Its Interaction with Explicit Cues in LMs. / Kementchedjhieva, Yova; Anderson, Mark; Søgaard, Anders.

Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021. Association for Computational Linguistics, 2021. p. 4859-4871.

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

Harvard

Kementchedjhieva, Y, Anderson, M & Søgaard, A 2021, John praised Mary because he ? Implicit Causality Bias and Its Interaction with Explicit Cues in LMs. in Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021. Association for Computational Linguistics, pp. 4859-4871, Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, Virtual, Online, 01/08/2021. https://doi.org/10.18653/v1/2021.findings-acl.429

APA

Kementchedjhieva, Y., Anderson, M., & Søgaard, A. (2021). John praised Mary because he ? Implicit Causality Bias and Its Interaction with Explicit Cues in LMs. In Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021 (pp. 4859-4871). Association for Computational Linguistics. https://doi.org/10.18653/v1/2021.findings-acl.429

Vancouver

Kementchedjhieva Y, Anderson M, Søgaard A. John praised Mary because he ? Implicit Causality Bias and Its Interaction with Explicit Cues in LMs. In Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021. Association for Computational Linguistics. 2021. p. 4859-4871 https://doi.org/10.18653/v1/2021.findings-acl.429

Author

Kementchedjhieva, Yova ; Anderson, Mark ; Søgaard, Anders. / John praised Mary because he ? Implicit Causality Bias and Its Interaction with Explicit Cues in LMs. Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021. Association for Computational Linguistics, 2021. pp. 4859-4871

Bibtex

@inproceedings{f3a09a5bdfe542d59e7b3e76bf12d4cf,
title = "John praised Mary because he ?: Implicit Causality Bias and Its Interaction with Explicit Cues in LMs",
abstract = "Some interpersonal verbs can implicitly attribute causality to either their subject or theirobject and are therefore said to carry an implicit causality (IC) bias. Through this bias,causal links can be inferred from a narrative,aiding language comprehension. We investigate whether pre-trained language models(PLMs) encode IC bias and use it at inferencetime. We find that to be the case, albeit todifferent degrees, for three distinct PLM architectures. However, causes do not alwaysneed to be implicit—when a cause is explicitlystated in a subordinate clause, an incongruentIC bias associated with the verb in the mainclause leads to a delay in human processing.We hypothesize that the temporary challengehumans face in integrating the two contradicting signals, one from the lexical semantics ofthe verb, one from the sentence-level semantics, would be reflected in higher error ratesfor models on tasks dependent on causal links.The results of our study lend support to this hypothesis, suggesting that PLMs tend to prioritize lexical patterns over higher-order signals.",
author = "Yova Kementchedjhieva and Mark Anderson and Anders S{\o}gaard",
year = "2021",
doi = "10.18653/v1/2021.findings-acl.429",
language = "English",
pages = "4859--4871",
booktitle = "Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021",
publisher = "Association for Computational Linguistics",
note = "Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021 ; Conference date: 01-08-2021 Through 06-08-2021",

}

RIS

TY - GEN

T1 - John praised Mary because he ?

T2 - Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021

AU - Kementchedjhieva, Yova

AU - Anderson, Mark

AU - Søgaard, Anders

PY - 2021

Y1 - 2021

N2 - Some interpersonal verbs can implicitly attribute causality to either their subject or theirobject and are therefore said to carry an implicit causality (IC) bias. Through this bias,causal links can be inferred from a narrative,aiding language comprehension. We investigate whether pre-trained language models(PLMs) encode IC bias and use it at inferencetime. We find that to be the case, albeit todifferent degrees, for three distinct PLM architectures. However, causes do not alwaysneed to be implicit—when a cause is explicitlystated in a subordinate clause, an incongruentIC bias associated with the verb in the mainclause leads to a delay in human processing.We hypothesize that the temporary challengehumans face in integrating the two contradicting signals, one from the lexical semantics ofthe verb, one from the sentence-level semantics, would be reflected in higher error ratesfor models on tasks dependent on causal links.The results of our study lend support to this hypothesis, suggesting that PLMs tend to prioritize lexical patterns over higher-order signals.

AB - Some interpersonal verbs can implicitly attribute causality to either their subject or theirobject and are therefore said to carry an implicit causality (IC) bias. Through this bias,causal links can be inferred from a narrative,aiding language comprehension. We investigate whether pre-trained language models(PLMs) encode IC bias and use it at inferencetime. We find that to be the case, albeit todifferent degrees, for three distinct PLM architectures. However, causes do not alwaysneed to be implicit—when a cause is explicitlystated in a subordinate clause, an incongruentIC bias associated with the verb in the mainclause leads to a delay in human processing.We hypothesize that the temporary challengehumans face in integrating the two contradicting signals, one from the lexical semantics ofthe verb, one from the sentence-level semantics, would be reflected in higher error ratesfor models on tasks dependent on causal links.The results of our study lend support to this hypothesis, suggesting that PLMs tend to prioritize lexical patterns over higher-order signals.

U2 - 10.18653/v1/2021.findings-acl.429

DO - 10.18653/v1/2021.findings-acl.429

M3 - Article in proceedings

SP - 4859

EP - 4871

BT - Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021

PB - Association for Computational Linguistics

Y2 - 1 August 2021 through 6 August 2021

ER -

ID: 300083155