Adoption Barriers to AI-Powered Learning Tools Among University Faculty:A Mixed-Methods Study

This mixed-methods study investigates the barriers that shape universityfaculty adoption of AI-powered learning tools, integrating a quantitative surveyof 412 faculty members across five disciplines with semi-structured interviews
from 18 purposively sampled participants. Machine learning techniques, including a random forest classifier and k-means clustering, were applied tothesurvey data to identify the strongest predictors of adoption and to segment
faculty into distinct adoption profiles. The random forest model classifiedfaculty adoption status with 72% accuracy and an area under the curve (AUC)
of 0.75, identifying perceived usefulness, perceived ease of use, and digital
literacy as the strongest predictors of adoption, while technology anxiety, limited institutional support, and workload pressure emerged as the most
salient barriers. Cluster analysis revealed three faculty segments, ReadyAdopters (34%), Cautious Adopters (37%), and Resistant Faculty (29%), eachrequiring differentiated support strategies. Qualitative interviews further
revealed five recurrent barrier themes: time and workload constraints, insufficient institutional training, concerns about academic integrity and dataprivacy, discomfort with algorithmic opacity, and a perceived mismatchbetween AI tools and disciplinary pedagogy. Supplementary robustness checks, including a random forest learning-curve analysis and a comparison against
logistic regression, confirmed the stability of the reported predictors across
modeling choices. The study concludes that faculty adoption of AI-poweredlearning tools is shaped less by the technology itself than by the organizational
and psychological ecosystem surrounding its introduction, and it proposes atiered, discipline-sensitive framework for institutional AI integration.