UX Design Principles for Trustworthy AI in Audiobook Services : An Ontology-Based Approach

Sammanfattning

Artificial intelligence systems are increasingly embedded in consumer-facing digital products, yet a significant gap remains between abstract ethical frameworks for Trustworthy AI and practical UX design guidance. This thesis addresses that gap through a user-centered study conducted in collaboration with Storytel, a digital audiobook streaming service. Five semi-structured interviews were analyzed using Grounded Theory, producing a taxonomy of 224 codes across 35 subcategories and eight main categories, with Transparency and Technical Robustness as the most valued dimensions. Through axial coding, subcategories were connected to Nielsen's ten usability heuristics using a weighted scoring method, yielding five UX/AI design principles that extend the heuristics to address AI-specific design challenges. These were formalized into an OWL ontology in Protégé and implemented as five high-fidelity prototypes, evaluated through a survey with 97 respondents. All mean scores of Likert-scale questions fell above the neutral midpoint, suggesting the principles are empirically grounded but require further implementation iteration.