Awareness of Generative AI Limitations and Students’ Intention to Use ChatGPT: A Moderated Mediation Analysis

Abstract

This research examined the importance of understanding generative artificial intelligence (GenAI) limitations in predicting college students’ likelihood of using ChatGPT, based on a modified Technology Acceptance Model. It treated awareness, perceived usefulness, and perceived anxiety about skills deterioration related to GenAI as independent variables, with intention to use ChatGPT as the dependent variable. The first-stage moderated mediation model was applied to analyze participants’ responses, with cooperation from 931 college students from different fields of study at a private higher education institution in the Philippines. The findings revealed that the awareness of GenAI limitations exerted a positive impact on the intention of college students to use ChatGPT, in particular, both its direct positive effect (c’ = 0.169, p < .001) and an indirect effect through the mediation of perceived usefulness. Conditional process analysis confirmed that skill-erosion concern moderated the relationship between awareness and perceived usefulness (moderated mediation index = -0.075, 95% CI [-0.107, -0.046]). Although perceived usefulness remained a significant mediator of the relationship between awareness and intention, the indirect effect diminished as skill-erosion concern increased, dropping from 0.233 for students with low-little concern to 0.105 for students with great concern. The collinearity analysis confirmed that the predictors were empirically distinguishable. The findings show that beliefs about risks regulate the technology acceptance process, and that AI-literacy training can be pursued alongside measured caution about dependency rather than treating these goals as alternatives.