Human-Centric Explanainable AI as a Catalyst for Responsible AI Adoption in Higher Education
Abstract
As Artificial Intelligence (AI) becomes increasingly integrated into academic workflows, the limited explainability of advanced AI models poses significant ethical and operational challenges, particularly in the Global South.This study proposes and validates an Extended Human-Centric Explainable AI (H-XAI) Adoption Framework within the context of Zimbabwean Higher Education. Utilizing a quantitative approach (N = 372 university students), regression analysis was employed to examine the relationships amongst five critical H-XAI constructs which are perceived interpretability, algorithm transparency, ethical alignment, AI literacy and human oversight together with traditional adoption drivers that include perceived usefulness, perceived ease of use, attitude and behavioral intention. Results demonstrate that the model explains 62% of the variance in Behavioral Intention (R2 = .62). Findings reveal that ethical alignment (β = .48, p < 0.001) and human oversight (β = .44, p < 0.001) are the strongest predictors of responsible AI adoption, while perceived interpretability (β=0.42, p < 0.001) significantly shapes Attitude. The implications for institutional governance and curriculum design are discussed, providing a roadmap for sustainable AI integration in resource-constrained educational environments.