Making Interaction Modes Visible in Human–AI Dialogue

Conversational AI systems dynamically adapt their behavior during dialogue, shifting between different interaction modes such as factual response, analytical processing, exploratory reasoning, or meta-level discussion about the interaction itself. While these shifts are often appropriate and beneficial, they generally remain implicit to the user. This opacity may lead to misunderstandings, loss of perceived control, or misinterpretation of the system’s intent. This note proposes the concept of an interaction mode indicator: an explicit, optional, and user-adjustable signal describing the interaction regime currently estimated by the system. Four broad regimes are distinguished—factual/technical, analytical/editorial, exploratory/heuristic, and conceptual/meta-reflective. Together, they illustrate how interaction transparency can support shared regulation and meta-collaborative cognition in human–AI dialogue, without anthropomorphism or user profiling.