Tracing the Path from Self-Efficacy to Artificial Intelligence Use: The Mediating Role of Technology Acceptance Among Mathematics Undergraduates in Nigeria

Abstract

The way students learn, engage with, and create information has changed as artificial intelligence (AI) becomes more common in classrooms. Therefore, understanding the factors that influence students' adoption of AI tools is vital to improving learning outcomes, especially in disciplines that require extensive mathematics. This study examined the mediating role of technology acceptance in the relationship between self-efficacy and AI use among mathematics undergraduates in Nigeria. We used a quantitative cross-sectional design and collected data from 256 students across three public universities in Southwestern Nigeria using a structured survey instrument. We operationalised technology acceptance solely as perceived ease of use, rather than as the full set of constructs typically represented in the Technology Acceptance Model. Data were analysed using PLS-SEM. The findings showed that self-efficacy significantly and positively relates to both technology acceptance (β = 0.591, t = 9.166, p < .05, f² = 0.537) and AI use (β = 0.301, t = 4.727, p < .05, f² = 0.092). Technology acceptance also significantly relates to AI use (β = 0.371, t = 5.093, p < .05, f² = 0.140) and partly mediated the relationship between self-efficacy and AI use. The model explained 35.0% of the variance in technology acceptance (R² = 0.350) and 36.1% of the variance in AI use (R² = 0.361), while both endogenous constructs demonstrated predictive relevance (Q² > 0). The findings suggest that mathematics educators can support AI use by strengthening students’ technological self-efficacy and ensuring that AI tools are easy to use within mathematics learning activities. Theoretically, the findings indicate that perceived ease of use provides an important mechanism through which self-efficacy relates to students’ AI use, while also highlighting the need to distinguish this specific acceptance dimension from broader technology acceptance models.