Strategies and approaches on explainable artificial intelligence

Abstract

This research thesis focuses on the problem of AI explainability. In recent years, the field of Explicable AI (XAI) has developed an extensive collection of algorithms that provide a useful set of tools for developers to create XAI applications. Explainability is regarded to have exceeded the demand for academics or scientists to comprehend the models they develop and has become a crucial requirement for consumers to embrace and trust AI in a variety of fields. This study describes existing XAI methods, pointing out their features and specific applications. In addition, it shows practical examples of the application of the described methods with real-world examples, pointing out the benefits of the results obtained. As a valuable outcome, a summary of the described methods with recommendations for their use is presented, as well as instructions on the basic steps necessary to integrate XAI into an existing business or a project based on the AI model.