Uncovering non-linearities in airline service experience: a dual cognitive-affective evaluation framework

This study proposes an explainable Machine Learning (ML) approach to unravel the non-linear, asymmetric, and threshold-dependent drivers of air passenger experience. By training parallel eXtreme Gradient Boosting (XGBoost) models paired with Shapley Additive exPlanations (SHAP), this research overcomes the limitations of traditional linear modelling, bridging dual-processing satisfaction theory and explainable AI (XAI) to disentangle analytical cognitive quality assessments from holistic affective emotional states. The findings of the proposed work reveal that service factors act differently across these two psychological domains. While operational factors like punctuality act as rigid hygiene factors for the rational mind, experiential aspects like boarding undergo a behavioural change, scaling as the primary accelerator of affective delight. Furthermore, SHAP interaction matrices identify asymmetric moderation effects driven by passenger expectations across different airline business models. By disentangling cognitive quality from affective delight, this research offers a novel eXplainable Artificial Intelligence (XAI) framework that helps airlines safeguard basic operational services against dissatisfaction while prioritizing key interactions that truly foster passenger delight.