<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="chhoda-2026-harm">
    <titleInfo>
        <title>When Harm Is Polite: Coercion Markers and Low Toxicity in Conversion-Therapy Discourse</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Adhiraj</namePart>
        <namePart type="family">Chhoda</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-10</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 10th Workshop on Online Abuse and Harms (WOAH)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Agostina</namePart>
            <namePart type="family">Calabrese</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Thomas</namePart>
            <namePart type="family">Davidson</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Christine</namePart>
            <namePart type="family">de Kock</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Urja</namePart>
            <namePart type="family">Khurana</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Marta</namePart>
            <namePart type="family">Marchiori Manerba</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Paloma</namePart>
            <namePart type="family">Piot</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Zeerak</namePart>
            <namePart type="family">Talat</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Budapest, Hungary</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-487-3</identifier>
    </relatedItem>
    <abstract>Ten years of online-harms research have largely operationalized harm through abusive tone: slurs, hostility, and platform-visible toxicity. Yet some harms are carried by polite, institutional, role-mediated discourse that standard toxicity classifiers consistently miss. We study one such case through a corpus of 2,388 publicly available conversion therapy (CT) documents (119,976 sentences) compared with 26 APA-aligned affirmative-therapy controls (16,948 sentences). We test a semantic moralization hypothesis using contextual association tests; the signed statistic (d = -5.11, p = 0.0002) indicates that identity terms in CT are on average more similar to neutral-psychology attributes than to morality attributes, contradicting the hypothesis. A transparent rule set based on the BITE model of coercive control finds all four dimensions elevated in CT, with emotional control highest. We test two further alternative mechanisms—a staged emotional script and grammatical agency-deletion—and find neither at corpus scale. Three toxicity classifiers (Detoxify, ToxiGen-HateBERT, ToxiGen-RoBERTa) do not register the elevated coercion markers: Detoxify and RoBERTa flag \ensuremathłeq1.4% of CT sentences; HateBERT flags 11.5% but triggers on identity-group vocabulary rather than coercive structure. These findings argue for online-harms taxonomies that incorporate relational context and semantic constraint alongside surface abuse, and we release derived features and code.</abstract>
    <identifier type="citekey">chhoda-2026-harm</identifier>
    <location>
        <url>https://aclanthology.org/2026.woah-1.5/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>38</start>
            <end>48</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T When Harm Is Polite: Coercion Markers and Low Toxicity in Conversion-Therapy Discourse
%A Chhoda, Adhiraj
%Y Calabrese, Agostina
%Y Davidson, Thomas
%Y de Kock, Christine
%Y Khurana, Urja
%Y Marchiori Manerba, Marta
%Y Piot, Paloma
%Y Talat, Zeerak
%S Proceedings of the 10th Workshop on Online Abuse and Harms (WOAH)
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-487-3
%F chhoda-2026-harm
%X Ten years of online-harms research have largely operationalized harm through abusive tone: slurs, hostility, and platform-visible toxicity. Yet some harms are carried by polite, institutional, role-mediated discourse that standard toxicity classifiers consistently miss. We study one such case through a corpus of 2,388 publicly available conversion therapy (CT) documents (119,976 sentences) compared with 26 APA-aligned affirmative-therapy controls (16,948 sentences). We test a semantic moralization hypothesis using contextual association tests; the signed statistic (d = -5.11, p = 0.0002) indicates that identity terms in CT are on average more similar to neutral-psychology attributes than to morality attributes, contradicting the hypothesis. A transparent rule set based on the BITE model of coercive control finds all four dimensions elevated in CT, with emotional control highest. We test two further alternative mechanisms—a staged emotional script and grammatical agency-deletion—and find neither at corpus scale. Three toxicity classifiers (Detoxify, ToxiGen-HateBERT, ToxiGen-RoBERTa) do not register the elevated coercion markers: Detoxify and RoBERTa flag \ensuremathłeq1.4% of CT sentences; HateBERT flags 11.5% but triggers on identity-group vocabulary rather than coercive structure. These findings argue for online-harms taxonomies that incorporate relational context and semantic constraint alongside surface abuse, and we release derived features and code.
%U https://aclanthology.org/2026.woah-1.5/
%P 38-48

Markdown (Informal)

[When Harm Is Polite: Coercion Markers and Low Toxicity in Conversion-Therapy Discourse](https://aclanthology.org/2026.woah-1.5/) (Chhoda, WOAH 2026)

• When Harm Is Polite: Coercion Markers and Low Toxicity in Conversion-Therapy Discourse (Chhoda, WOAH 2026)

ACL

• Adhiraj Chhoda. 2026. When Harm Is Polite: Coercion Markers and Low Toxicity in Conversion-Therapy Discourse. In Proceedings of the 10th Workshop on Online Abuse and Harms (WOAH), pages 38–48, Budapest, Hungary. Association for Computational Linguistics.