Multi-criteria decision model for AI readiness in academic libraries

This study aims to develop and test a multi-criteria decision model for assessing the preparedness of academic libraries for AI adoption in higher education. Using input from 21 head librarians in Poland, we identified 25 key indicators of AI readiness. Probabilistic weights at both domain and indicator levels were estimated using a hierarchical Bayesian Best–Worst Method, with partial-order priorities expressed through credal rankings. A fuzzy VIKOR synthesis produced a compromise readiness ranking for five institutional archetypes. Domain priorities emphasize AI’s impact on staff roles (0.219) and its use as a tool (0.209), followed by current AI use (0.194), assistant functions (0.189), and reversed AI-as-threat (0.189). Sensitivity analysis reveals a tipping point: when the weight for current use exceeds approximately 0.30, practice-and-skill indicators surpass adoption concerns. Fuzzy VIKOR analysis ranked polytechnic libraries as the most prepared for AI adoption in higher education institutions, followed by research libraries.