Empowered or Constrained? Digital Agency, Ethical Implications, and Students’ Intentions to Use Artificial Intelligence

Drawing on digital agency theory, expectancy–value frameworks, and self-regulated learning perspectives, this study proposes and tests a moderated mediation model explaining students’ intentions to use AI. Using data from 673 university students, we examined whether sense of positive agency (SOPA) predicts intention to use AI indirectly through perceived value and perceived benefits of AI, and whether these pathways are conditionally shaped by sense of negative agency (SONA). Conditional process analysis (PROCESS Model 59) showed that SOPA had no direct effect on intention to use AI (b = 0.013, p = 0.882). Instead, its influence was fully indirect and conditional. SOPA predicted perceived value and perceived benefits of AI only at moderate to high levels of SONA, with significant SOPA × SONA interactions for both mediators (p = 0.040). Perceived value strongly predicted intention to use AI (b = 0.385, p < 0.001), and this relationship was amplified at higher levels of negative agency (b = 0.138, p = 0.002). In contrast, the effect of perceived benefits on intention weakened as SONA increased (b = −0.125, p = 0.005), becoming non-significant at higher levels of negative agency (Johnson–Neyman point ≈ 2.99). The final model explained 50.4% of the variance in intention to use AI. Overall, the findings indicate a conditional appraisal mechanism: as negative agency increases, perceived value becomes a stronger predictor of intention, whereas the motivational contribution of perceived benefits weakens and becomes non-significant beyond the Johnson–Neyman threshold. These results support an agency-aware account of AI adoption focused on how cognitive appraisals relate to intention under different perceived agency orientations, without implying ethical reasoning or moral deliberation processes not measured in this study.