User Profiles and Perceptions for Predicting Acceptance with Fairness Assessment in Border Control Technologies

Abstract In 2013, European border crossing points (BCPs) introduced border control technologies (BCTs) aimed at improving efficiency and security. However, these technologies face challenges in user adoption and acceptance, often due to limited traveler awareness and support at BCPs. Traditional models like the technology acceptance model (TAM) rely on structured data and struggle with complex, non-linear relationships, limiting their effectiveness in predicting technology acceptance. Recent advancements in machine learning offer enhanced accuracy for predicting technology acceptance by leveraging unstructured, large-scale data from sources like online social networks (OSNs) or questionnaires. Existing studies have focused primarily on individual attributes like age and gender, with limited attention to how these attributes interact with user perceptions. Despite these advancements, research on technology acceptance prediction for BCTs remains limited, particularly in integrating user profile attributes and perceptions and also no comprehensive relevant dataset exists for this purpose. This study addresses this gap by using an automated approach for developing technology acceptance score dataset that combines user profiles and perception attributes along with investigating how the integration of these attributes impact prediction accuracy and assess its fairness by considering bias and explainable AI (XAI) in the developed models. Our results indicate that this integrated approach significantly enhances the prediction of technology acceptance, with TabTransformer model achieving the highest accuracy of 76%. This study not only contributes to the existing research domain of technology acceptance prediction but also opens new directions for future research into technology acceptance for BCTs as well as for other types of technologies.