Research Agenda for Basic Explainable AI

Artificial Intelligence is increasingly driven by powerful but often opaque machine learning algorithms. These black-box algorithms achieve high performance but are not explainable to humans in a systematic and interpretable manner, a challenge known as Explainable AI (XAI). Informed by a synthesis of two converging literature streams on information systems development and psychology, we propose a new XAI approach termed Basic Explainable AI and a subsequent research agenda. We propose four research directions that focus on providing explanations by proactively considering the target audience's mental models and making the explanations maximally accessible to heterogeneous nonexpert users.