Answers are grounded in this paper and cite their sources.
Universities need to know which factors raise students’ intention to use generative artificial intelligence and which must be present before high intention can occur. The hybrid structural equation and neural network analyses that dominate this literature address only the first question, rank predictors with a heuristic importance index and rarely test prediction on new respondents. We re-analyzed an open survey of 1,242 ChatGPT-using university students from 78 Philippine institutions, combining partial least squares structural equation modeling, necessary condition analysis, gradient boosting explained with Shapley values and cross-validated predictive tests. Habit was the strongest unique predictor of intention to use ChatGPT yet ranked sixth or seventh of seven on necessity under every scoring, whereas effort expectancy, with no detectable unique effect, ranked among the top three. Necessity magnitudes depended on the Likert response distributions and fell below conventional thresholds after rank transformation, so the robust result is this inverted ordering rather than the magnitudes. The three importance rankings agreed at the top and bottom, and neither the structural model nor the boosted ensemble predicted new respondents better than a linear benchmark. Separating predictors that raise intention from those that must be present gives institutions a clearer basis for sequencing support.
No takes yet. Share an insight, caveat, or question.
Nguyen et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: