Research-Thesis: Ethical customisation in AI travel booking: a between-subjects experiment on perceived agency, trust, and satisfaction

Abstract

This thesis explores whether giving travellers more control over the ethical settings of AI recommendation systems can improve their satisfaction. Using a between-subjects experiment with 122 participants, respondents were presented with either a standard Booking.com-style interface or a version that included visible ethical preference controls related to eco-friendliness, data privacy, and social impact. The findings showed that the ethical sliders themselves did not directly increase satisfaction. Instead, their effect came from the greater sense of control they provided to the users. Perceived agency emerged as a significant mediator, suggesting that participants responded more positively when they felt actively involved in shaping the recommendation process. In contrast, platform trust and perceived ethical alignment did not significantly mediate the relationship, possibly because these perceptions develop over time through repeated interactions rather than from a single exposure. Additionally, exploratory factor analysis revealed limited discriminant validity between these three constructs in a single-exposure context. The study also found that users with higher levels of algorithmic literacy were generally more satisfied with AI-powered recommendation systems, although they also appeared more critical of superficial ethical signs. Overall, the results suggest that ethical customisation features are most valuable when they enhance users’ sense of autonomy rather than simply serving as ethical branding. For platform designers, the central implication is that making ethical controls visible and clearly labelled, even before users interact with them, is a low-cost intervention with meaningful psychological impact.