GUI Design Patterns for Improving the HCI in Explainable Artificial Intelligence

The rising number of artificial intelligence systems has increased the demand for transparency and accountability. This thesis addresses the human-computer interaction (HCI) challenges of explainable artificial intelligence (XAI), focusing on reusable solutions for transparency and fairness challenges. The main objectives we aim to address are 1) Identification of the most common and most significant HCI usability challenges in XAI; 2) Identification of good and reusable solutions for recognised HCI challenges in literature and existing XAI solutions, and 3) Their definition in the form of the design patterns. A catalogue of graphical user interface design patterns (i.e. reusable solutions to commonly occurring problems with a given context) in XAI is proposed. Expected benefits of the thesis include facilitating communication and collaboration between researchers and practitioners by providing shared vocabulary and proven design patterns, enhancing trust and accountability of AI solutions, promoting user-centred design and facilitating the evaluation and accelerating the maturation of XAI.