Comparative analysis of similarity levels between texts generated by ChatGPT-3.5 and human texts through Turnitin Software evaluation

 

Bewaard in:
Bibliografische gegevens
Auteur: Obando Zambrano, Beyker Stefan
Formaat: artículo original
Status:Versión publicada
Publicatiedatum:2024
Omschrijving:The research focused on identifying the degree of similarity between documents generated by the generative artificial intelligence ChatGPT-3.5 and those created without its assistance, comparing them with the documents available in the Turnitin software databases. To achieve this objective, a descriptive research study with a quantitative approach was conducted, evaluating a total of 300 documents: 150 created entirely by ChatGPT-3.5 and 150 without the use of technological tools. These documents were produced by students from the University of Costa Rica, Guanacaste Campus, and uploaded to the Moodle platform in a space enabled with Turnitin software during the first semester of 2023. The results showed that the documents generated by ChatGPT-3.5 had an average similarity level of 22% with respect to the Turnitin system databases, significantly lower compared to the documents created without its intervention, which had an average similarity level of 33%. Additionally, the collected data indicate that, whether using generative artificial intelligence or not, it is essential to exercise caution in properly referencing documents to avoid plagiarism.
Land:Portal de Revistas UCR
Instelling:Universidad de Costa Rica
Repositorio:Portal de Revistas UCR
Taal:Español
OAI Identifier:oai:portal.ucr.ac.cr:article/57267
Online toegang:https://revistas.ucr.ac.cr/index.php/eciencias/article/view/57267
Keyword:Generative artificial intelligence
information technologies
similarity
documents
Inteligencia artificial generativa
tecnologías de información
similitud
documentos
Inteligência artificial generativa
plágio
tecnologias de informação
competências digitais