How trust prospectively predicts perceived fairness and learning motivation toward AI teaching assistants: a three-wave within-person longitudinal study

Artificial intelligence teaching assistants (AITAs) are increasingly embedded in higher education, yet how students’ trust, fairness perceptions, and motivation form and reciprocally shape one another over time remains poorly understood, because cross-sectional designs cannot separate stable between-person differences from genuine within-person change. This three-wave prospective longitudinal study ( N = 450 undergraduates; 4-week intervals across an 8-week semester) used a random-intercept cross-lagged panel model (RI-CLPM)—with the standard CLPM as a sensitivity benchmark—to examine within-person relationships among student trust (ability, benevolence, integrity), perceived fairness (distributive, procedural, interactional justice), and learning motivation (intrinsic, extrinsic). Longitudinal measurement invariance held across waves. Within-person increases in trust predicted subsequent increases in perceived fairness, with ability trust emerging as the strongest antecedent. Procedural justice showed a within-person association with subsequent intrinsic motivation that was numerically larger than its association with extrinsic motivation, although a direct test of this coefficient difference did not itself reach conventional significance; bias-corrected bootstrapping confirmed a significant longitudinal mediation linking trust to motivation through fairness. Reciprocal effects from fairness back to trust were substantially weaker than trust-to-fairness effects, indicating an asymmetric rather than strictly unidirectional pattern of influence. These findings point to early ability-trust cultivation and procedural transparency as candidate targets for psychologically grounded, human-centered AITA design, although an observational design cannot establish that intervening on them would produce the corresponding changes.