Contemporality
Communication property of receiving information as it is produced with minimal delay.
Contemporality is a property of communication identified in Clark and Brennan's grounding framework: participants receive information as it is produced by others, with minimal delay. In synchronous face-to-face conversation, information is available to all participants immediately as it is produced — there is no buffering, no transmission latency, no asynchronous delivery. The speaker's pauses, reformulations, and emphases are all available to the listener in real time, allowing the listener to adjust their interpretation continuously rather than having to reassess a complete message after it has been fully produced.
The value of contemporality for collaborative work is that it allows participants to provide real-time feedback, clarification, and coordination without waiting for the other party to finish. A listener who is confused by a speaker's phrasing can signal that confusion mid-sentence (through facial expression, a backchannel utterance, or an interjection), and the speaker can adjust — either by reformulating immediately or by explicitly addressing the confusion — before the original formulation has been completed. This reduces the amount of rework and miscommunication that occurs when complete messages turn out to be misunderstood.
In AI interaction, contemporality requires that the model perceive and respond to user inputs with minimal latency — not processing complete turns before responding, but updating its understanding continuously as the input unfolds. The interaction model's time-aligned micro-turn architecture is specifically designed to provide this: by processing 200ms chunks of audio and video as they arrive, the model maintains a contemporaneous representation of the user's current contribution, allowing it to respond to partial inputs in a way that turn-based systems cannot.
The practical challenge is that full contemporality is computationally expensive. Processing partial inputs requires the model to reason over incomplete information — a harder problem than reasoning over complete, finalized inputs. The interaction model's architecture handles this by maintaining a continuously updated state that represents the current partial input, which the background reasoning model can query. This separation between the contemporaneous interaction layer and the deeper reasoning layer is what makes contemporality tractable at the quality level required for useful collaboration.