Quando as máquinas parecem humanas demais: intrusividade, estranheza e cinismo da privacidade no uso de chatbots de IA generativa

Abstract

This dissertation examines the cognitive antecedents of privacy cynicism in automated interactions, with a particular focus on perceived intrusiveness and perceived strangeness in chatbot use. Additionally, it explores the mediating role of perceived vulnerability in the relationship between privacy cynicism and intentions to protect privacy. This study adopted a quantitative research design. It was conducted via an online survey of 257 Brazilian users who had interacted with AI-based chatbots in their daily activities within 6 months of data collection. Data were examined using Partial Least Squares Structural Equation Modeling (PLS-SEM). Perceived intrusiveness exerted a more substantial effect on privacy cynicism than perceived strangeness, although both constructs were significant predictors. In addition, perceived vulnerability fully mediated the relationship between privacy cynicism and privacy protection intention, indicating that cynicism alone does not necessarily lead to protective behavior. The findings offer guidance for the design of AI-powered customer service systems, emphasizing the importance of balancing personalization, transparency, and data protection in ways that are consistent with users’ perceptions and expectations. This study highlights the societal importance of developing technologies that safeguard user autonomy, dignity, and trust in digital environments. The findings contribute to broader debates on ethical AI and data protection, supporting the advancement of inclusive and transparent governance practices that help mitigate digital fatigue and privacy resignation. Keywords: