This thesis explores opportunities for enhancing the adoption of Artificial Intelligence (AI) in real-world credit card fraud detection, focusing on Explainable Artificial Intelligence (XAI). XAI aims to improve the transparency of complex AI models by providing explanations for their outcomes. Although XAI is an emerging field, it is increasingly regarded by regulatory bodies, public organizations, and private corporations as crucial for the future of AI and automated decision-making systems.

The thesis begins with a broad perspective, introducing the domain of credit card fraud detection through an overview of the operating environment and recent regulatory and technological changes driving AI adoption. It then provides a survey of existing literature to propose a research agenda supporting advancements in XAI for fraud detection practices.

As the thesis progresses, the next chapter delves into more detail regarding one particular element of the research agenda, namely ensuring the context of the explanation is taken into account. This chapter establishes common ground by providing definitions of key concepts in normative XAI contexts. It then introduces the SAGE (Setting, Audience, Goals, and Ethics) framework, which consolidates recent discourse to create a user-centric XAI solution.

The practical implementation of the SAGE framework is demonstrated through a real-world scenario in the subsequent chapter. At this point the reader is introduced to the principles of Scenario-Based Design (SBD) and the concept of functionally grounded evaluation which, together with the SAGE framework, provide an XAI solution which more accurately represents real-world requirements than established XAI models.

Overall, this structured approach ensures a thorough understanding of the topic, transitioning from a broad overview to detailed examinations of key elements. The thesis advances knowledge on the contextual aspects important for building real-world XAI solutions for credit card fraud detection and demonstrates how the SAGE framework can be leveraged for this purpose.