Without a transcript there is no way to tell whether the system is listening, thinking or about to speak, so people talk over it and lose their input.
← Back to patterns
Toward Sociable Robots Cost-Based Anticipatory Action Selection for Human–Robot Fluency
Agent Workflow
EmergingMachine State, Rendered
The Channel That Is Left When There Is No Transcript
When the interface has no transcript to read, the system’s state has to live in a non-textual, pre-attentive channel — a glyph, a ring, a colour, a motion. Stage Tracker answers how far along a task is; this answers whether it is safe to speak now, on a sub-second clock instead of a minutes-long one. The states that earn it are the ones a transcript cannot express: heard-you-but-still-listening, barge-in accepted, muted, both-at-once. Mature in the field since Alexa’s light ring in 2014 — novel only to this archive.
Metaphor
Light Ring
Interaction
Hold-to-Talk, Barge-In, Mute
Use case
Voice Assistants, Ambient Devices, Hands-Free Interfaces, Live Agents
Control mechanisms
State Glyph · Named Channel States · Timed Event Log
Reference examples
Amazon Alexa light ring · ChatGPT Advanced Voice Mode · Gemini Live
Interactive Example
The live side-by-side is the comparison.
This prototype already contains the contrast between chat and the pattern response. It stands as the primary comparison surface for this page.
The indicator said Speaking the entire time.
You start talking at 1,100 ms, stop at 3,500. 2,400 ms of your audio reaches the mic and zero characters reach the transcript, because the one indicator on screen read “Speaking…” throughout — whether the system was hearing you or not. The turn ends, the text is written at 4,400 ms, and your half is simply missing.
Nine named states in a channel that needs no reading, and five of them are ones no transcript can carry: heard-you-but-still-listening, both-at-once, barge-in accepted, muted, channel error. Beside the glyph, every transition names the machine event that produced it — gain gate, ASR partial, endpointer, first token, TTS buffer — with millisecond stamps.
Stage Tracker answers how far along a task is, on a clock measured in minutes. This answers whether it is safe to speak now, on a clock measured in tenths of a second. Full-duplex models made that question live again: when a system can listen and speak at once, turn ownership stops being obvious, and silence is no longer a reliable signal for either party.
A progress bar tells you to wait. A turn-taking signal tells you when not to.
DOPE Lens
How this pattern improves AI interaction quality.
Observability
A closed set of named states renders continuously in a channel that needs no reading, so turn ownership is visible at a glance rather than inferred from silence
Directability
Barge-in is a first-class state rather than an error: speaking over the assistant produces an acknowledged transition, not a collision
Predictability
The states are declared and finite, and each one names the machine event that produces it — endpoint detected, first token, buffer playing
Explainability
An event log timestamps every transition beside the glyph, so a state that looks wrong can be traced to the signal that caused it
Fit
Where it earns its place.
Avoid
when
when
The interface already shows a transcript and turn boundaries are legible from the messages themselves.
In the wild
Where this pattern already ships.
The pattern is shipping but still maturing across products. Entries below show how different teams have approached it; expect variation as the space settles.
Reference implementations
Amazon Alexa light ring · ChatGPT Advanced Voice Mode · Gemini Live
Detailed write-ups for each implementation are in progress.
Research Frame
Why this pattern exists, and where it breaks.
Render a small closed set of channel states in a non-textual, pre-attentive surface, and timestamp every transition beside it.
Chat resolves turn-taking with a visible message boundary; strip the transcript and that boundary disappears, leaving silence to carry the whole signal.
A glyph that only changes colour is a mood ring — if every state maps onto a phase label a text tracker could print, this is a progress display wearing a face.
Common questions
What people ask about this pattern.
When should I avoid the Machine State, Rendered pattern?
+
The interface already shows a transcript and turn boundaries are legible from the messages themselves.
What problem does Machine State, Rendered solve?
+
Without a transcript there is no way to tell whether the system is listening, thinking or about to speak, so people talk over it and lose their input.
Why is chat the wrong fit for this?
+
Chat resolves turn-taking with a visible message boundary; strip the transcript and that boundary disappears, leaving silence to carry the whole signal.
Grounded in Research
Not just opinion.
The papers, guidelines, and usability studies this pattern rests on.
Cynthia Breazeal
MIT Media Lab · Robotics and Autonomous Systems, Vol. 42, Iss. 3–4, pp. 167–175 · 2003
Names the para-linguistic envelope display — gaze, brow, lean — as the channel that hands the speaking floor back and forth, and shows it works on timing rather than content: the same silent pause reads as “still mine” or “yours now” depending only on where the robot looks. Kismet held about 82.5% clean turns with four naive subjects and no shared transcript.
Guy Hoffman, Cynthia Breazeal
MIT Media Lab · IEEE Transactions on Robotics, Vol. 23, Iss. 5, pp. 952–961 · 2007
Separates perceived fluency from task efficiency: with outcomes statistically indistinguishable, subjects rated the anticipatory robot far higher on commitment to the team (5.0 vs 2.8, p<0.005), and the only measurable differences were temporal. Grounds the clock, not the channel — cite it for timing being what gets read, never for sub-second thresholds.
Related Patterns
Curated connections, not just neighbors.
Hand-picked relationships — what this pattern pairs with in practice, what it replaces, what it builds on, and what builds on it.
Often paired with
Read-Along
Machine State, Rendered says whose turn it is; Read-Along adds which word of the spoken answer is playing.
Often paired with
Mid-Stream Steering
Mid-Stream Steering redirects an answer while it runs; Machine State, Rendered shows the barge-in was accepted, not lost.
Alternative to
Stage Tracker
Stage Tracker tracks a long task's phases over minutes; Machine State, Rendered shows whether it is safe to speak right now.