Measuring Student Trust and Over-Reliance on AI Tutors: Implications for STEM Learning Outcomes

Abstract

The introduction of artificial intelligence (AI) tutors into science, technology, engineering, and mathematics (STEM) education has changed the way learners acquire and digest information. Although the use of AI tutors promotes individualized education, it also conveys significant questions of how much students trust technology and how over-dependence they have become on the automated systems. This paper examines how trust in AI tutors is correlated with learning behaviours of the students in terms of the effects of the levels of trust on cognitive engagement, autonomy in solving problems, and learning outcomes. The mixed-methods research design was used, involving structured surveys, behavioral usage data, and semi-structured interviews with undergraduate STEM learners in different institutes. The correlation between the levels of trust and the signs indicating over-reliance was studied through quantitative analysis, whereas the results of the qualitative study gave the findings about the perceptions of students towards the capabilities of AI tutors. Findings indicate that moderate trust has the potential to improve learning effectiveness and confidence but over-trust decreases critical thinking and independent problem-solving. The research acknowledges that the future of human-AI interaction should focus on balanced patterns of interaction encouraging the development of trust without affecting the autonomy of the learners. They give recommendations that educators, designers, and policymakers can implement to achieve the responsible use of AI in STEM education.