RELATIONAL COGNITION, MORAL RESPONSIBILITY, AND TRUST IN CONVERSATIONAL AI: EVIDENCE FROM MONGOLIA

Abstract

This study investigates how culturally embedded relational cognition shapes user interaction and trust formation in conversational artificial intelligence as a digital information processing environment. While prior research on human-AI interaction has largely assumed universal patterns of social response, little attention has been given to how culturally specific cognitive orientations influence interaction style, trust development, and perceived user autonomy. Drawing on survey data from 327 Mongolian users of ChatGPT, this study conceptualizes relational cognition as a multidimensional orientation encompassing indirect communication, face management, respect for social hierarchy, emotional sensitivity, and relationally delayed decision-making.

Grounded in cultural psychology, intercultural communication theory, and information processing research, the study examines three questions: (1) how relational cognition predicts interaction style with conversational AI; (2) how relational cognition and interaction style shape trust in AI systems; and (3) whether trust in AI is associated with diminished or preserved human agency. Using exploratory factor analysis, multiple regression modelling, and bootstrapped mediation analysis, the findings demonstrate that relational cognition significantly predicts polite, indirect, and emotionally sensitive interaction styles. These culturally patterned interactional behaviors partially mediate the relationship between relational cognition and trust in AI. Importantly, neither relational cognition nor trust is associated with reduced perceived human agency. Instead, users exhibiting stronger relational orientations report reflective and autonomous engagement with AI-generated information.

The study contributes to information processing and management literature in four ways. First, it extends models of trust in digital systems by conceptualizing trust as culturally mediated rather than purely technical.

Second, it introduces relational cognition as a meaningful construct for understanding culturally situated information interaction. Third, it reconceptualizes the Computers Are Social Actors paradigm as culturally differentiated rather than universal. Fourth, it provides empirical evidence from Mongolia, an underrepresented developing-country context in AI research.

The findings highlight the importance of culturally sensitive design and governance of conversational AI systems and caution against universal assumptions about automation bias and loss of autonomy. Understanding AI-mediated information environments therefore requires attention not only to system performance but also to the cultural cognitive frameworks through which users interpret and engage with technology.