Self-regulated motivation in AI-mediated L2 speaking practice: the roles of AI literacy, autonomy, and self-perceived competence

Grounded in Self-Determination Theory (SDT), this study investigated the relationship between AI literacy and self-regulated speaking motivation among 152 Macao English major undergraduates using partial least squares structural equation modeling (PLS-SEM). Results showed that AI literacy was positively related to autonomy (β = .470, p < .001), self-perceived competence (β = .452, p < .001), and task-value activation (β = .373, p < .001), but not directly with environment regulation (β = .062, p = .515). Significant indirect relationships involving autonomy were observed between AI literacy and both task-value activation (β = .196, p = .005) and environment regulation (β = .209, p < .001). In contrast, a significant indirect relationship involving self-perceived competence was observed between AI literacy and environment regulation (β = .145, p < .001), but not task-value activation (β = .022, p = .680). The model explained 22.1% of the variance in autonomy, 20.5% in self-perceived competence, 50.4% in task-value activation, and 51.5% in environment regulation. The findings suggest that AI literacy was not directly associated with environment regulation after autonomy and self-perceived competence were considered, whereas significant indirect relationships involving these psychological resources were observed. It provides implications for designing autonomy-supporting AI-mediated speaking practice.