Without visible feedback, silence could be a hang or a normal wait

What it is

After speaking, if the screen shows neither listening nor processing nor a result—only silence—people cannot tell a system that did not respond (hung, never heard, network down) from a normal wait (endpoint still waiting, server still computing). Missing visible feedback stacks two opposite next steps—try again / wait a bit more—into one experience.

Why it happens

Tolerance for unacknowledged waits is short, and people bias toward the worst attribution. Voice also lacks touch’s instantaneous receipt. Silence is therefore an ambiguous signal: VAD still waiting for you to finish, a request in queue, a permission dialog on another layer, a crash. A visible state labels the wait as one of these; a further timeout becomes an explicit failure. Without that, repeated speech, tapping empty space, and giving up happen together. With no continuous listen, no level, and no processing state, what remains is this ambiguous silence.

Studying it

Under a legitimately long wait and a true failure (drop the network, kill the recognizer), hide all status UI and log when people respeak, leave, and what they blame. Then turn status UI back on as a control. Dependent measures include false respeaking and premature abandonment. An experimenter in the room makes people more willing to wait; the field is harsher.

Where it stops holding

On a familiar short skill (“what time is it”) people will wait a beat in empty air. On first use and on high-cost actions, silence is unacceptable. An auditory comfort tone can replace visible feedback when eyes-free. Labelling every silence an error immediately false-alarms on a normal half-second of processing.

Applying it

• Any blank past budget must carry a named state; a further timeout becomes a failure with a next step.
• Failure copy should distinguish “didn’t hear / network / cannot complete,” not one generic bang.
• Accept with a dropped network and an over-long process, and confirm there are no silent seconds with nothing on screen.

Related

• Same group: C7.16.1 While speaking, continuous visual feedback must prove the system is capturing · C7.16.2 Level or waveform motion lets people tell the mic is receiving sound, not that the device has failed · C7.16.3 After recognition, mark a shift into processing, distinct from still-listening
• Adjacent: C7.09 Endpoint Detection and End-of-Utterance · C7.10 Live Partial Results
• Search: dead air · wait ambiguity · voice timeout