TRUST MATTERS: ONGOING STAKEHOLDER CONFIDENCE IN AI FOR EDUCATION

Abstract

Adoption of Artificial Intelligence in educational settings is gaining momentum, with the trust of stakeholders becoming an important issue. This research develops a machine learning-based model to predict the trust of stakeholders in AI in education. We collected 1,113 survey responses from participants, engineering 35 features on demographic attributes, AI-related behaviors, and perceptions. Our best model, a Logistic Regression classifier, yields 86.23% accuracy and an AUC-ROC of 0.927 in predicting high versus low trust. Its performance in 5-fold cross-validation is 82.35%. These results show that predictors of significance are perceptions of the future of AI and concerns, while the main pattern of behavior is the usage frequency. This research offers a validated data-driven framework for understanding and predicting AI trust in education. Along with offering insights into how institutions could foster confidence in AI technologies, the paper positions its contribution within the state-of-the-art landscape.