I'm not a robot: análise semiótica de interações entre humanos e o chatbot Tay

• Ano de defesa: 2022
• Autor(a) principal: Ludmilla Teixeira Lima
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; Brasil; FALE - FACULDADE DE LETRAS; Programa de Pós-Graduação em Estudos Linguísticos; UFMG
• Palavras-chave em Português: Semiótica Discursiva; chatbot Tay; Microsoft; Inteligência Artificial; enunciador bipartido; Semiótica; Linguística; Inteligência artificial
• Link de acesso: http://hdl.handle.net/1843/45763; https://orcid.org/0000-0002-2134-6276

The aim of this work is to semiotically investigate part of the linguistic interactions of the chatbot Tay (launched by Microsoft in 2016) performed on the online platform Twitter as a support for its technological existence. In this context, we seek to answer the following question: how do human interaction written utterances with chatbot Tay are discursively constructed, considering enunciative strategies and their meaning effects. Our theoretical-methodological scope is that of Discursive Semiotics (also called French or Greimasian), based on the readings of Greimas and Courtés (2021), Greimas (2014), as well as its promoters in Brazil, such as Barros (1990, 2002, 2005), Lara and Matte (2009a and 2009b) and Fiorin (2000), among others. Some concepts from the field of Computer Science are briefly explained to support the reader in understanding our object of study. Focusing on the study of the discursive level (actarial organization, themes and figures) and the oppositions of the fundamental level, we analysed six interactions between Tay and users who, through generally offensive messages, tried altering the robot's purpose. We conclude that, as for the enunciative organisation, specifically at the level of discursive syntax, the use of artificial intelligence technology in the construction of the robot generates a “bipartite enunciator” between the company that created it and the user base that took control of its speech by imposing the offensive language Tay learnt from the interactions. Semantically, the analysed utterances/discourses concretise figurative dominance paths, supported using iconization processes, anchoring and reality effects according to the theme imposed by the enunciators-users. At a deeper level, the oppositions life x death and identity x alterity, abstractly show which categories the discourse is based on. We also concluded that Discursive Semiotics did contribute to the understanding of human-machine interaction, by revealing the semiotic existence of a chatbot.