This paper introduces PREF-XAI, a novel approach for generating personalized, rule-based explanations of black-box machine learning models. It reframes explanation as a preference-driven decision problem, learning individual user preferences through robust ordinal regression to tailor explanations.