Generative AI acceptance in study-travel contexts: an integrative model of affective and utilitarian pathways with mediation and grade-level moderation.

BACKGROUND: Generative artificial intelligence is progressively reshaping study-travel-a curriculum-embedded educational practice. Yet, the psychological processes governing upper-primary and middle-school students' adoption of such technology in this context remain underexplored. Prior technology acceptance inquiries have predominantly centered on adult populations, leaving adolescent cognitive-developmental contingencies unaddressed.

OBJECTIVE: This investigation aimed to elucidate the differential roles of affective versus utilitarian drivers in shaping study-travel participants' anticipated intentions to adopt GenAI, and to determine whether these pathways differ by grade level.

METHOD: We constructed an integrative model grounded in the extended Unified Theory of Acceptance and Use of Technology (UTAUT2), supplemented by Piaget's developmental framework. The model incorporated perceived enjoyment as a mediator and perceived risk as a direct negative antecedent, with grade level specified as a moderator of selected pathways. Survey data from 741 students in grades 4-9, drawn from twelve schools across six Chinese cities, were analyzed using structural equation modeling, multi-group comparisons, and cluster-robust standard errors to account for the nested sampling structure. Measurement invariance across grade groups was established prior to structural comparisons.

RESULTS: Perceived enjoyment emerged as the largest estimated predictor of adoption intention in this sample (β = 0.43, p < 0.001), surpassing performance expectancy (β = 0.30, p < 0.001). Enjoyment fully mediated social influence effects and partially transmitted performance expectancy's influence. Perceived risk exerted a significant suppressive effect (β = -0.24, p < 0.001), with primary students exhibiting approximately twice the inhibitory magnitude of middle-school counterparts. Grade significantly moderated the performance expectancy-intention relationship, and provided borderline evidence for moderation of the risk-intention relationship, indicating systematic grade-related differences in how these pathways operate.

CONCLUSION: These findings advance technology acceptance theory by introducing grade-related contingency and establishing an "affective-efficacy" dual-channel model. Practical implications for designing grade-differentiated educational AI products are discussed.