@inproceedings{jeoung-etal-2023-examining,
title = "Examining the Causal Impact of First Names on Language Models: The Case of Social Commonsense Reasoning",
author = "Jeoung, Sullam and
Diesner, Jana and
Kilicoglu, Halil",
editor = "Ovalle, Anaelia and
Chang, Kai-Wei and
Mehrabi, Ninareh and
Pruksachatkun, Yada and
Galystan, Aram and
Dhamala, Jwala and
Verma, Apurv and
Cao, Trista and
Kumar, Anoop and
Gupta, Rahul",
booktitle = "Proceedings of the 3rd Workshop on Trustworthy Natural Language Processing (TrustNLP 2023)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.trustnlp-1.7",
doi = "10.18653/v1/2023.trustnlp-1.7",
pages = "61--72",
abstract = "As language models continue to be integrated into applications of personal and societal relevance, ensuring these models{'} trustworthiness is crucial, particularly with respect to producing consistent outputs regardless of sensitive attributes. Given that first names may serve as proxies for (intersectional) socio-demographic representations, it is imperative to examine the impact of first names on commonsense reasoning capabilities. In this paper, we study whether a model{'}s reasoning given a specific input differs based on the first names provided. Our underlying assumption is that the reasoning about Alice should not differ from the reasoning about James. We propose and implement a controlled experimental framework to measure the causal effect of first names on commonsense reasoning, enabling us to distinguish between model predictions due to chance and caused by actual factors of interest. Our results indicate that the frequency of first names has a direct effect on model prediction, with less frequent names yielding divergent predictions compared to more frequent names. To gain insights into the internal mechanisms of models that are contributing to these behaviors, we also conduct an in-depth explainable analysis. Overall, our findings suggest that to ensure model robustness, it is essential to augment datasets with more diverse first names during the configuration stage.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="jeoung-etal-2023-examining">
<titleInfo>
<title>Examining the Causal Impact of First Names on Language Models: The Case of Social Commonsense Reasoning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Sullam</namePart>
<namePart type="family">Jeoung</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jana</namePart>
<namePart type="family">Diesner</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Halil</namePart>
<namePart type="family">Kilicoglu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2023-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 3rd Workshop on Trustworthy Natural Language Processing (TrustNLP 2023)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Anaelia</namePart>
<namePart type="family">Ovalle</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kai-Wei</namePart>
<namePart type="family">Chang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ninareh</namePart>
<namePart type="family">Mehrabi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yada</namePart>
<namePart type="family">Pruksachatkun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aram</namePart>
<namePart type="family">Galystan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jwala</namePart>
<namePart type="family">Dhamala</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Apurv</namePart>
<namePart type="family">Verma</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Trista</namePart>
<namePart type="family">Cao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anoop</namePart>
<namePart type="family">Kumar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rahul</namePart>
<namePart type="family">Gupta</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Toronto, Canada</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>As language models continue to be integrated into applications of personal and societal relevance, ensuring these models’ trustworthiness is crucial, particularly with respect to producing consistent outputs regardless of sensitive attributes. Given that first names may serve as proxies for (intersectional) socio-demographic representations, it is imperative to examine the impact of first names on commonsense reasoning capabilities. In this paper, we study whether a model’s reasoning given a specific input differs based on the first names provided. Our underlying assumption is that the reasoning about Alice should not differ from the reasoning about James. We propose and implement a controlled experimental framework to measure the causal effect of first names on commonsense reasoning, enabling us to distinguish between model predictions due to chance and caused by actual factors of interest. Our results indicate that the frequency of first names has a direct effect on model prediction, with less frequent names yielding divergent predictions compared to more frequent names. To gain insights into the internal mechanisms of models that are contributing to these behaviors, we also conduct an in-depth explainable analysis. Overall, our findings suggest that to ensure model robustness, it is essential to augment datasets with more diverse first names during the configuration stage.</abstract>
<identifier type="citekey">jeoung-etal-2023-examining</identifier>
<identifier type="doi">10.18653/v1/2023.trustnlp-1.7</identifier>
<location>
<url>https://aclanthology.org/2023.trustnlp-1.7</url>
</location>
<part>
<date>2023-07</date>
<extent unit="page">
<start>61</start>
<end>72</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Examining the Causal Impact of First Names on Language Models: The Case of Social Commonsense Reasoning
%A Jeoung, Sullam
%A Diesner, Jana
%A Kilicoglu, Halil
%Y Ovalle, Anaelia
%Y Chang, Kai-Wei
%Y Mehrabi, Ninareh
%Y Pruksachatkun, Yada
%Y Galystan, Aram
%Y Dhamala, Jwala
%Y Verma, Apurv
%Y Cao, Trista
%Y Kumar, Anoop
%Y Gupta, Rahul
%S Proceedings of the 3rd Workshop on Trustworthy Natural Language Processing (TrustNLP 2023)
%D 2023
%8 July
%I Association for Computational Linguistics
%C Toronto, Canada
%F jeoung-etal-2023-examining
%X As language models continue to be integrated into applications of personal and societal relevance, ensuring these models’ trustworthiness is crucial, particularly with respect to producing consistent outputs regardless of sensitive attributes. Given that first names may serve as proxies for (intersectional) socio-demographic representations, it is imperative to examine the impact of first names on commonsense reasoning capabilities. In this paper, we study whether a model’s reasoning given a specific input differs based on the first names provided. Our underlying assumption is that the reasoning about Alice should not differ from the reasoning about James. We propose and implement a controlled experimental framework to measure the causal effect of first names on commonsense reasoning, enabling us to distinguish between model predictions due to chance and caused by actual factors of interest. Our results indicate that the frequency of first names has a direct effect on model prediction, with less frequent names yielding divergent predictions compared to more frequent names. To gain insights into the internal mechanisms of models that are contributing to these behaviors, we also conduct an in-depth explainable analysis. Overall, our findings suggest that to ensure model robustness, it is essential to augment datasets with more diverse first names during the configuration stage.
%R 10.18653/v1/2023.trustnlp-1.7
%U https://aclanthology.org/2023.trustnlp-1.7
%U https://doi.org/10.18653/v1/2023.trustnlp-1.7
%P 61-72
Markdown (Informal)
[Examining the Causal Impact of First Names on Language Models: The Case of Social Commonsense Reasoning](https://aclanthology.org/2023.trustnlp-1.7) (Jeoung et al., TrustNLP 2023)
ACL