The Implications of Generative-AI User Prompt Engineering Shaping the Absent Presence of Latino Women Textual Rhetoric in LLMs
Year: 2026
Type: article
Abstract: This article examines how large language models (LLMs) generate text that affects underrepresented groups, particularly Latino women, due to limited cultural, contextual, and abstraction training rooted in biases. The research draws on Brock’s (2018) CTDA theory, White et al.’s (2023) prompt engineering framework, Klein’s (2015) computational spatiality, and Anzaldua’s (1987) concepts of cultural, emotional, and gendered borderlands to analyze the outputs. Using narrative prompt experiments, Chat GPT 3.5, 5, Claude AI, Gemini, and Microsoft Co-Pilot produced changes in their textual outputs as evidence of further algorithmic designs. Nevertheless, continual biases embedded in the algorithms’ linguistic representation and misrepresentation through the absence of contextual and abstraction interpretations, and profession stereotypes persisted. Despite the technologies’ design advancement, evidence of the limitations of Latina characters were consistently represented through tropes of immigration, Spanglish use, and food consumption, and generational cultural transference in the kitchen spatiality of the home. I argue that LLMs training relies on cultural, political, social, and economic worldviews that have historically been rooted in harmful biases while new data is shaped by biases in general. The study aims to enhance LLMs' algorithmic effect in shaping the perception of cultural and its limitations to remove bias through continued research and development. Hence, this research benefits program designers, educators, writers, and students.
Author: Suarez DeJesus, Angely C
Institution: University of Central Florida
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7165245562
