Comparing Implementations of Explainable Artificial Intelligence using End-User Evaluations: Protocol for a Systematic Review

In this protocol, we outline the steps for a systematic review on different implementations of Explainable Artificial Intelligence (XAI) evaluated by end-users. We aim to answer the research question: How do different implementations of XAI influence human-AI interaction? Therefore, we search for studies that compare different implementations of XAI to each other on human-AI interaction outcomes, such as user-perception measures and collaboration performance measures. We aim to create an overview of empirical studies that involve humans in the evaluation of XAI and compare different XAI implementations to each other, to gain more understanding on the factors contributing to a good explanation of AI systems and how this could be potentially different between application domains.