Answers are grounded in this paper and cite their sources.
Human–AI conversations are increasingly evolving from simple question–answer exchanges into continuous, interactive, multimodal, and potentially multi-party communication environments. However, conventional conversational systems predominantly rely on turn-taking mechanisms, end-of-turn detection, automatic speech recognition, and interruption handling to determine when a participant can speak. These mechanisms do not adequately distinguish between directly interrupting an ongoing speaker and formally requesting permission to contribute while that speaker is still communicating. This paper introduces Generative AI Interruption Request (GAI-IR), a novel conceptual framework that treats interruption requests as an independent and explicitly manageable conversational action between humans and generative artificial intelligence systems. Unlike conventional barge-in mechanisms, which allow one participant to interrupt another, GAI-IR establishes a request-oriented interaction process in which a participant can express an intention to contribute, supplement, clarify, correct, question, or provide additional information during another participant's ongoing utterance. The framework supports bidirectional human-to-AI and AI-to-human requests, unidirectional configurations, and extensible multi-party configurations. Its proposed capabilities include request initiation, recognition, notification, permission granting, refusal, postponement, queuing, prioritization, withdrawal, expiration, partial authorization, scheduled intervention, contextual evaluation, and post-intervention conversational restoration. The framework further distinguishes the right to request an interruption from the right to execute an interruption, thereby establishing a separation between conversational intention and conversational authorization. To the best of our knowledge, this paper introduces the first unified conceptual framework specifically centered on bidirectional and unidirectional interruption requests between humans and generative AI systems, rather than merely detecting, executing, or recovering from interruptions. The paper establishes the conceptual foundations, interaction taxonomy, architectural components, operational principles, evaluation dimensions, and research propositions necessary for developing interruption-request-aware conversational systems.
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: