Publikation:

How user language affects conflict fatality estimates in ChatGPT

Lade...
Vorschaubild

Dateien

Zu diesem Dokument gibt es keine Dateien.

Datum

2024

Autor:innen

Steinert, Christoph Valentin

Herausgeber:innen

Kontakt

ISSN der Zeitschrift

Electronic ISSN

ISBN

Bibliografische Daten

Verlag

Schriftenreihe

Auflagebezeichnung

URI (zitierfähiger Link)
ArXiv-ID

Internationale Patentnummer

Angaben zur Forschungsförderung

Projekt

Open Access-Veröffentlichung
Open Access Hybrid
Core Facility der Universität Konstanz

Gesperrt bis

Titel in einer weiteren Sprache

Publikationstyp
Zeitschriftenartikel
Publikationsstatus
Published

Erschienen in

Journal of Peace Research. Sage. ISSN 0022-3433. eISSN 1460-3578. Verfügbar unter: doi: 10.1177/00223433241279381

Zusammenfassung

OpenAI’s ChatGPT language model has gained popularity as a powerful tool for problem-solving and information retrieval. However, concerns arise about the reproduction of biases present in the language-specific training data. In this study, we address this issue in the context of the Israeli–Palestinian and Turkish–Kurdish conflicts. Using GPT-3.5, we employed an automated query procedure to inquire about casualties in specific airstrikes, in both Hebrew and Arabic for the former conflict and Turkish and Kurdish for the latter. Our analysis reveals that GPT-3.5 provides 34 ± 11% lower fatality estimates when queried in the language of the attacker than in the language of the targeted group. Evasive answers denying the existence of such attacks further increase the discrepancy. A simplified analysis on the current GPT-4 model shows the same trends. To explain the origin of the bias, we conducted a systematic media content analysis of Arabic news sources. The media analysis suggests that the large-language model fails to link specific attacks to the corresponding fatality numbers reported in the Arabic news. Due to its reliance on co-occurring words, the large-language model may provide death tolls from different attacks with greater news impact or cumulative death counts that are prevalent in the training data. Given that large-language models may shape information dissemination in the future, the language bias identified in our study has the potential to amplify existing biases along linguistic dyads and contribute to information bubbles.

Zusammenfassung in einer weiteren Sprache

Fachgebiet (DDC)
530 Physik

Schlagwörter

armed conflict, artificial intelligence, ChatGPT, conflict fatalities, indiscriminate violence, large-language models

Konferenz

Rezension
undefined / . - undefined, undefined

Forschungsvorhaben

Organisationseinheiten

Zeitschriftenheft

Zugehörige Datensätze in KOPS

Zitieren

ISO 690STEINERT, Christoph Valentin, Daniel L. KAZENWADEL, 2024. How user language affects conflict fatality estimates in ChatGPT. In: Journal of Peace Research. Sage. ISSN 0022-3433. eISSN 1460-3578. Verfügbar unter: doi: 10.1177/00223433241279381
BibTex
@article{Steinert2024-11-03langu-71396,
  year={2024},
  doi={10.1177/00223433241279381},
  title={How user language affects conflict fatality estimates in ChatGPT},
  issn={0022-3433},
  journal={Journal of Peace Research},
  author={Steinert, Christoph Valentin and Kazenwadel, Daniel L.}
}
RDF
<rdf:RDF
    xmlns:dcterms="http://purl.org/dc/terms/"
    xmlns:dc="http://purl.org/dc/elements/1.1/"
    xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
    xmlns:bibo="http://purl.org/ontology/bibo/"
    xmlns:dspace="http://digital-repositories.org/ontologies/dspace/0.1.0#"
    xmlns:foaf="http://xmlns.com/foaf/0.1/"
    xmlns:void="http://rdfs.org/ns/void#"
    xmlns:xsd="http://www.w3.org/2001/XMLSchema#" > 
  <rdf:Description rdf:about="https://kops.uni-konstanz.de/server/rdf/resource/123456789/71396">
    <dc:creator>Steinert, Christoph Valentin</dc:creator>
    <foaf:homepage rdf:resource="http://localhost:8080/"/>
    <dc:contributor>Kazenwadel, Daniel L.</dc:contributor>
    <dc:language>eng</dc:language>
    <dspace:isPartOfCollection rdf:resource="https://kops.uni-konstanz.de/server/rdf/resource/123456789/41"/>
    <dcterms:issued>2024-11-03</dcterms:issued>
    <dcterms:abstract>OpenAI’s ChatGPT language model has gained popularity as a powerful tool for problem-solving and information retrieval. However, concerns arise about the reproduction of biases present in the language-specific training data. In this study, we address this issue in the context of the Israeli–Palestinian and Turkish–Kurdish conflicts. Using GPT-3.5, we employed an automated query procedure to inquire about casualties in specific airstrikes, in both Hebrew and Arabic for the former conflict and Turkish and Kurdish for the latter. Our analysis reveals that GPT-3.5 provides 34 ± 11% lower fatality estimates when queried in the language of the attacker than in the language of the targeted group. Evasive answers denying the existence of such attacks further increase the discrepancy. A simplified analysis on the current GPT-4 model shows the same trends. To explain the origin of the bias, we conducted a systematic media content analysis of Arabic news sources. The media analysis suggests that the large-language model fails to link specific attacks to the corresponding fatality numbers reported in the Arabic news. Due to its reliance on co-occurring words, the large-language model may provide death tolls from different attacks with greater news impact or cumulative death counts that are prevalent in the training data. Given that large-language models may shape information dissemination in the future, the language bias identified in our study has the potential to amplify existing biases along linguistic dyads and contribute to information bubbles.</dcterms:abstract>
    <dc:date rdf:datatype="http://www.w3.org/2001/XMLSchema#dateTime">2024-11-22T08:24:22Z</dc:date>
    <dc:contributor>Steinert, Christoph Valentin</dc:contributor>
    <dcterms:available rdf:datatype="http://www.w3.org/2001/XMLSchema#dateTime">2024-11-22T08:24:22Z</dcterms:available>
    <dcterms:rights rdf:resource="http://creativecommons.org/licenses/by/4.0/"/>
    <dc:rights>Attribution 4.0 International</dc:rights>
    <void:sparqlEndpoint rdf:resource="http://localhost/fuseki/dspace/sparql"/>
    <dcterms:title>How user language affects conflict fatality estimates in ChatGPT</dcterms:title>
    <dcterms:isPartOf rdf:resource="https://kops.uni-konstanz.de/server/rdf/resource/123456789/41"/>
    <bibo:uri rdf:resource="https://kops.uni-konstanz.de/handle/123456789/71396"/>
    <dc:creator>Kazenwadel, Daniel L.</dc:creator>
  </rdf:Description>
</rdf:RDF>

Interner Vermerk

xmlui.Submission.submit.DescribeStep.inputForms.label.kops_note_fromSubmitter

Kontakt
URL der Originalveröffentl.

Prüfdatum der URL

Prüfungsdatum der Dissertation

Finanzierungsart

Kommentar zur Publikation

Allianzlizenz
Corresponding Authors der Uni Konstanz vorhanden
Internationale Co-Autor:innen
Universitätsbibliographie
Ja
Begutachtet
Ja
Online First: Zeitschriftenartikel, die schon vor ihrer Zuordnung zu einem bestimmten Zeitschriftenheft (= Issue) online gestellt werden. Online First-Artikel werden auf der Homepage des Journals in der Verlagsfassung veröffentlicht.
Diese Publikation teilen