A Communicative Action Framework for Discourse Strategies for AI-based Systems: The MetTrains Application Case

Increasing attention is being paid to the challenges of how artificial intelligence (AI)-based systems offer explanations to users. Explanation capabilities developed for older logic-based systems still have relevance, but new thinking is needed in designing explanations and other discourse strategies for new forms of AI that include machine learning. In this work-in-progress paper we show how a communicative action design framework can be used to design an AI-based system’s interface to achieve desired goals. The applicability of the framework is demonstrated with an interface for an intelligent video surveillance system for reducing railway suicide. The communicative action framework is an important step in theory development for human-computer interaction with AI as used in the information systems domain.