In a library or a sleeping room, the question isn't whether a sound can be heard but whether it should be

What it is

In a library, a bedroom at night, or a quiet meeting room, the obstacle to letting a device make a sound is usually not "can it be heard" but "should it be heard." The quieter the environment, the higher the SNR of any audio output relative to that environment by default — meaning the sound will be noticed easily and abruptly, by the user and by everyone nearby, at the cost of interrupting others, exposing private content, or making the user look inconsiderate. This is a different constraint from intelligibility: an intelligibility problem is a sound too weak to hear; a social-cost problem is a sound that, once loud enough to be heard, causes trouble beyond simply being heard.

Why it happens

A low ambient noise floor means the minimum SNR needed for intelligibility is trivially satisfied — the sound is inherently "loud enough" without any extra boost. But the flip side of that same physical fact is that the sound just as easily clears the detection threshold of every non-target listener nearby, turning it into an uncontrolled broadcast to everyone present rather than something confined to the user's own field of view the way visual output can be. Sound has no directional occlusion, and a quiet environment amplifies the consequences of that indiscriminate propagation — notification content, a caller's identity, a message fragment can all be overheard by others in the same room, adding a layer of privacy exposure and social awkwardness on top of the interruption itself.

Where it stops holding

• The cost is situational, not a property of the sound itself: the same alert tone goes almost unnoticed in a noisy café and draws complaints in a bedroom at night — the judgment is set by situational norms, not a decibel number.
• Private wearable listening devices (headphones, bone conduction) can satisfy the user's own intelligibility needs without changing the ambient quietness of the room, so this constraint doesn't apply to them; it only applies to audio meant to play openly into the environment.
• Whether a setting counts as "quiet" can flip quickly within the same time and place (the same meeting room is quiet during a session and noisy during a coffee break), so a static context category cannot cover every case — it has to be judged against the current state.

Applying it

• In a setting known or inferable to be quiet (late night, a calendar showing an ongoing meeting, ambient noise measured as low for a sustained period), default output to silent plus haptic or visual feedback, rather than defaulting to sound and leaving the user to mute it manually.
• For urgent notifications that cannot be delayed (a call, a safety alert), keep sound as a fallback, but lead with a lower-key transitional cue (a brief single chime, a fade-in rather than an abrupt burst) that gives the user a chance to confirm manually or switch to a private channel before the full content plays at full volume.
• How to verify it: walk the device through real quiet settings and log whether each audio output shifts the attention of bystanders or triggers avoidance behavior from the user (lowering volume, answering quickly, leaving the room) — that behavior is itself the signal that a social cost has already occurred.

Related

• Same group: A3.09.1 SNR, not absolute noise level, determines intelligibility · A3.09.2 speech input and output degrade together in noise · A3.09.4 the SNR needed for intelligibility varies with content type and familiarity · A3.09.5 ambient noise floors swing sharply across contexts, defeating fixed volume settings · A3.09.6 active noise cancellation changes what reaches the ear, not the device's own output loudness setting
• Nearby: A3.02 loudness perception and equal-loudness contours
• Search terms: acoustic etiquette · notification disturbance · public soundscape · context-aware audio