Aplicación de herramientas de Inteligencia Artificial como metodología emergente para el análisis de la toxicidad de la conversación en redes sociales
No Thumbnail Available
Identifiers
Publication date
2024-04
Defense date
2024-02-14
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
Se estudia una nueva metodología de inteligencia artificial (IA), entendiendo que la comunicación se presenta como uno de los campos de trabajo más trascendentes para su aplicación. Además de las fases de recolección y producción de contenido, otras áreas dentro del mundo de la comunicación como la distribución, y en concreto la moderación de comentarios -en plataformas sociales y en medios- también están viviendo un período de innovación, pero de forma menos evidente de cara a la audiencia. Por ello, en este trabajo de investigación se procede a conocer cómo el ensamblaje entre diversas herramientas de IA (Communalytic, Perspective API y Botometer) puede medir la calidad de la conversación y combatir la toxicidad en espacios comunicativos. Para ello, se han analizado 125,551 tweets publicados entre el 1 y el 15 de abril de 2023 correspondientes a quince políticos de España, Francia e Italia y a la cascada de contestaciones recibidas de sus audiencias. De esta manera, las herramientas de IA permiten conocer en detalle las tasas de diversas categorías de toxicidad y la tipología de la naturaleza de los usuarios que conversan. En consecuencia, se han abierto varias líneas de discusión sobre la generación de toxicidad y el incremento de bots según factores geográficos, institucionales y de género. Igualmente, se percibe una diferencia de comportamiento en los usuarios entre la conversación horizontal entre pares y la vertical con los políticos. Visto lo anterior, esta investigación muestra cómo una nueva metodología de IA ayuda a visibilizar y cuantificar nuevas realidades como la toxicidad online, con el fin último de llegar a erradicarla y sanear el debate en el mundo digital. Además, contradice los hallazgos previos sobre bots como difusores de toxicidad, demostrando que son los usuarios reales quienes amplifican el contenido tóxico.
A new artificial intelligence (AI) methodology is studied, with the understanding that communication is one of the most important fields of work for its application. In addition to the content collection and production phases, other areas within the world of communication such as distribution, and specifically the moderation of comments - on social platforms and in the media - are also experiencing a period of innovation, but in a less obvious way in the eyes of the audience. For this reason, in this research work we proceed to find out how the assembly of various AI tools (Communalytic, Perspective API and Botometer) can measure the quality of the conversation and combat toxicity in communicative spaces. To this end, 125,551 tweets published between 1st and 15th April 2023 corresponding to fifteen politicians from Spain, France and Italy and the cascade of replies received from their audiences were analysed. In this way, AI tools allow us to know in detail the rates of various categories of toxicity and the typology of the nature of the users who converse. As a result, several lines of discussion have been opened on the generation of toxicity and the increase of bots according to geographical, institutional and gender factors. Likewise, there is a perceived difference in user behaviour between horizontal conversation between peers and vertical conversation with politicians. In light of the above, this research shows how a new AI methodology helps to make visible and quantify new realities such as online toxicity, with the ultimate goal of eradicating it and cleaning up the debate in the digital world. Moreover, it contradicts previous findings on bots as disseminators of toxicity, demonstrating that it is real users who amplify toxic content.
A new artificial intelligence (AI) methodology is studied, with the understanding that communication is one of the most important fields of work for its application. In addition to the content collection and production phases, other areas within the world of communication such as distribution, and specifically the moderation of comments - on social platforms and in the media - are also experiencing a period of innovation, but in a less obvious way in the eyes of the audience. For this reason, in this research work we proceed to find out how the assembly of various AI tools (Communalytic, Perspective API and Botometer) can measure the quality of the conversation and combat toxicity in communicative spaces. To this end, 125,551 tweets published between 1st and 15th April 2023 corresponding to fifteen politicians from Spain, France and Italy and the cascade of replies received from their audiences were analysed. In this way, AI tools allow us to know in detail the rates of various categories of toxicity and the typology of the nature of the users who converse. As a result, several lines of discussion have been opened on the generation of toxicity and the increase of bots according to geographical, institutional and gender factors. Likewise, there is a perceived difference in user behaviour between horizontal conversation between peers and vertical conversation with politicians. In light of the above, this research shows how a new AI methodology helps to make visible and quantify new realities such as online toxicity, with the ultimate goal of eradicating it and cleaning up the debate in the digital world. Moreover, it contradicts previous findings on bots as disseminators of toxicity, demonstrating that it is real users who amplify toxic content.
Description
Mención Internacional en el título de doctor
Keywords
Toxicidad, Inteligencia Artificial, Bots, Metodología, Moderación, Redes sociales, Medios de comunicación, Toxicity, Artificial Intelligence, Bots, Methodology, Moderation, Social media, Media outlets