In our modern society, software systems are highly integrated into our daily life and will continue to be so. They execute tasks that were accomplished by specially trained humans before and often have to interact with human staff. Quality aspects such as ethics, fairness, and transparency have been discussed as essential for trustworthy software systems. Explainability has been identified as not only a means to achieve all of these three in systems, but also as a way to actually foster users’ sentiments of trust in systems. Like other quality aspects, explainability must be discovered and treated during the design of those systems. Although explainability has become a hot topic and communities from different areas of knowledge (e.g., machine learning, HCI, philosophy, psychology, cyber-physical and recommender systems) have been researching it actively, there is only little research on systematic explainability engineering. Yet, we believe that methods and techniques from requirements and software engineering would add a lot of value to the explainability research.
As a first step to explore this research landscape, we held an interdisciplinary workshop to collect ideas from different communities and to discuss open research questions. In a subsequent working group, we further analyzed and structured the results of this workshop to identify the most important research questions. As a result, we now present a research roadmap for explainable systems. With our research roadmap we aim to advance the software engineering methods and techniques for explainable systems and to attract research on the most urgent open questions.