From Explainable AI to Interactive AI: Conditions for a Paradigm Shift in Human–Computer Interaction and Design Theory

This article does not assume that interactive AI is the inevitable successor to explainable AI. Instead, it asks what kinds of interaction materially change users’ control over goals, outputs, and system rules. Through a critical synthesis of selected literature and a mechanism-oriented comparison of five research and product cases, it distinguishes three overlapping research strands: technical explainability, human-centered explanation, and participatory or collaborative AI. It proposes a nested feedback model, a context-sensitive hierarchy of interaction capability, and a four-dimensional evaluation framework covering cognition, action, relations, and justice. The article’s theoretical contribution is to define a “paradigm shift” as a joint reorganization of the design object, temporal structure, intervention rights, and knowledge formation, rather than a change from an explanation panel to a chat interface. The model and indicators are theoretical propositions that remain to be tested and have not yet been validated through cross-product user studies.