Banca de QUALIFICAÇÃO: JULIANA RODRIGUES DE LIMA

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
STUDENT : JULIANA RODRIGUES DE LIMA
DATE: 03/09/2025
TIME: 15:00
LOCAL: Google Meet
TITLE:

Fusion Between Human and Machine: An Epistemological Investigation of AI Concepts in Academic Discourse in the Light of Cognitive Linguistics

KEY WORDS:

Cognitive Linguistics; Conceptual Metaphor; Conceptual Blending; Artificial Intelligence.

PAGES: 88
BIG AREA: Lingüística, Letras e Artes
AREA: Lingüística
SUBÁREA: Teoria e Análise Lingüística
SUMMARY:

Although Artificial Intelligence (AI) is widely studied from technical and applied perspectives, little is known about how experts in the field construct conceptual meanings to define it in their scientific discourse. This theoretical-epistemological gap hinders a deeper understanding of the cognitive schemas that underpin knowledge production in AI, especially in the context of contemporary generative models, potentially leading to misconceptions about how these systems actually function. Thus, the general objective of this research is to characterize the concepts that make up the domain of Artificial Intelligence (AI) based on the theoretical assumptions of Cognitive Linguistics, focusing on how these concepts are cognitively modeled by specialists in the field. This characterization aims to reveal the symbolic and linguistic mechanisms that structure the understanding of AI in academic discourse in order to address the theoretical-epistemological gap. To investigate this gap, a qualitative, descriptive study was conducted, focusing on the characterization of AI concepts through the lens of Cognitive Linguistics. The research corpus consisted of 14 scientific articles selected from the Web of Science database, published between 2020 and 2024, that address generative AI models. The analysis focused on verbal cues in the texts, using specialized reading techniques and textual analysis tools such as Voyant Tools to identify linguistic and cognitive patterns in the representation of AI. The variables investigated included conceptual domains, metaphors (Lakoff & Johnson, 1980), and conceptual blends (Fauconnier & Turner, 2002) underlying the identified concepts. The analysis revealed the predominance of the conceptual domain MACHINE in the discourse of specialists, with recurring terms such as machine, model, system, data, and network, associated with the technical functioning of AI. Selective projections from the HUMAN domain were also identified, as evidenced by terms such as learn, train, understand, performance, and decision, suggesting conceptual approximations to human cognition. Three main metaphors were mapped — MIND AS MACHINE, AI AS AUTONOMOUS SYSTEM, and AI AS A SOURCE OF TRUTH — along with incomplete conceptual blends that combine technical and human domains without a full fusion between them. The research, grounded in Cognitive Linguistics, characterized how specialists model AI concepts through metaphors and selective conceptual integrations, partially and structurally projecting human attributes onto technical systems. These findings answer the research question by showing that meaning-making about AI in academic discourse is marked by hybrid schemas and conceptual tensions between the MACHINE and HUMAN domains. Moreover, these conceptual tensions revealed contradictions in AI epistemology, such as in the metaphor AI AS AUTONOMOUS SYSTEM, in which AI is presented as independent, while at the same time experts emphasize the presence of humans in the AI training process. It is worth noting that this contradiction also shapes our understanding of AI (Reddy, 1979), potentially leading to an overestimation of these systems in the face of a lack of transparency in discourse.

COMMITTEE MEMBERS:
Interno - 1675404 - PAULO HENRIQUE DUQUE
Externa à Instituição - ILANA SOUTO DE MEDEIROS - UnP
Externo à Instituição - RICARDO YAMASHITA SANTOS - UnP