Answers are grounded in this paper and cite their sources.
Generative artificial intelligence (AI) has rapidly transformed human–computer communication, enabling increasingly natural interactions through text, speech, and multimodal interfaces. Nevertheless, the temporal control of speaking behavior remains an important but insufficiently unified dimension of conversational interaction. Existing research has primarily investigated turn-taking, interruption handling, speech endpoint detection, conversational silence, and the suspension or resumption of system-generated speech. These mechanisms do not necessarily establish a general framework in which both human users and generative AI systems are independently recognized as speaking entities with explicit capabilities to temporarily suspend and subsequently resume their own contributions. This paper introduces the Bidirectional Speaking Pause Framework (BSPF), a conceptual framework for controlling temporary speaking suspension in generative AI-mediated communication. The framework distinguishes between AI-initiated and user-initiated speaking pauses, unidirectional and bidirectional suspension, independent and coordinated pause states, and the subsequent resumption, cancellation, or replacement of suspended speech. Unlike conventional interruption mechanisms, which primarily concern one participant interrupting another, BSPF conceptualizes speaking suspension as an independent communication operation that can be initiated by either participant without necessarily terminating the conversation or transferring the speaking right. The framework further establishes a separation between speaking state, listening state, conversational state, and content-generation state, allowing a participant to suspend speech while preserving the broader interaction context. To the best of the author's knowledge, BSPF represents the first proposed unified conceptual framework specifically integrating independently controlled human and generative-AI speaking pauses across both unidirectional and bidirectional interaction configurations. This contribution provides a foundation for future implementations, interaction protocols, and empirical investigations of temporal control in generative AI communication.
No takes yet. Share an insight, caveat, or question.
Jincheng Zhang (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: