<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="iskander-etal-2023-shielded">
    <titleInfo>
        <title>Shielded Representations: Protecting Sensitive Attributes Through Iterative Gradient-Based Projection</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Shadi</namePart>
        <namePart type="family">Iskander</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Kira</namePart>
        <namePart type="family">Radinsky</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yonatan</namePart>
        <namePart type="family">Belinkov</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>Findings of the Association for Computational Linguistics: ACL 2023</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Anna</namePart>
            <namePart type="family">Rogers</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jordan</namePart>
            <namePart type="family">Boyd-Graber</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Naoaki</namePart>
            <namePart type="family">Okazaki</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>Natural language processing models tend to learn and encode social biases present in the data. One popular approach for addressing such biases is to eliminate encoded information from the model’s representations. However, current methods are restricted to removing only linearly encoded information. In this work, we propose Iterative Gradient-Based Projection (IGBP), a novel method for removing non-linear encoded concepts from neural representations. Our method consists of iteratively training neural classifiers to predict a particular attribute we seek to eliminate, followed by a projection of the representation on a hypersurface, such that the classifiers become oblivious to the target attribute. We evaluate the effectiveness of our method on the task of removing gender and race information as sensitive attributes. Our results demonstrate that IGBP is effective in mitigating bias through intrinsic and extrinsic evaluations, with minimal impact on downstream task accuracy.</abstract>
    <identifier type="citekey">iskander-etal-2023-shielded</identifier>
    <identifier type="doi">10.18653/v1/2023.findings-acl.369</identifier>
    <location>
        <url>https://aclanthology.org/2023.findings-acl.369</url>
    </location>
    <part>
        <date>2023-07</date>
        <extent unit="page">
            <start>5961</start>
            <end>5977</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Shielded Representations: Protecting Sensitive Attributes Through Iterative Gradient-Based Projection
%A Iskander, Shadi
%A Radinsky, Kira
%A Belinkov, Yonatan
%Y Rogers, Anna
%Y Boyd-Graber, Jordan
%Y Okazaki, Naoaki
%S Findings of the Association for Computational Linguistics: ACL 2023
%D 2023
%8 July
%I Association for Computational Linguistics
%C Toronto, Canada
%F iskander-etal-2023-shielded
%X Natural language processing models tend to learn and encode social biases present in the data. One popular approach for addressing such biases is to eliminate encoded information from the model’s representations. However, current methods are restricted to removing only linearly encoded information. In this work, we propose Iterative Gradient-Based Projection (IGBP), a novel method for removing non-linear encoded concepts from neural representations. Our method consists of iteratively training neural classifiers to predict a particular attribute we seek to eliminate, followed by a projection of the representation on a hypersurface, such that the classifiers become oblivious to the target attribute. We evaluate the effectiveness of our method on the task of removing gender and race information as sensitive attributes. Our results demonstrate that IGBP is effective in mitigating bias through intrinsic and extrinsic evaluations, with minimal impact on downstream task accuracy.
%R 10.18653/v1/2023.findings-acl.369
%U https://aclanthology.org/2023.findings-acl.369
%U https://doi.org/10.18653/v1/2023.findings-acl.369
%P 5961-5977

Markdown (Informal)

[Shielded Representations: Protecting Sensitive Attributes Through Iterative Gradient-Based Projection](https://aclanthology.org/2023.findings-acl.369) (Iskander et al., Findings 2023)

• Shielded Representations: Protecting Sensitive Attributes Through Iterative Gradient-Based Projection (Iskander et al., Findings 2023)

ACL

• Shadi Iskander, Kira Radinsky, and Yonatan Belinkov. 2023. Shielded Representations: Protecting Sensitive Attributes Through Iterative Gradient-Based Projection. In Findings of the Association for Computational Linguistics: ACL 2023, pages 5961–5977, Toronto, Canada. Association for Computational Linguistics.