Public sound output exposes what the user is doing

What it is

Sound output leaks content in public. A screen can be angled so only the front sees it; sound radiates outward with no way to limit the audience. Playing specific content tells strangers what the user is doing, who it involves, and what state they are in.

Why it happens

Severity depends on three things: whether the level exceeds ambient noise, how specific the content is, and how well bystanders can infer. Even a short cue lets those nearby infer a conversation partner, a schedule, or a health state. Unlike vision, sound has no physical boundary; headphones mitigate, but unplugging, speakerphone, or external playback restores the leak instantly. Leakage also accumulates: individually harmless cues together sketch a full activity profile.

Studying it

Study audibility range and inferential reach in situated settings: place participants at varying distances and angles and record what they can hear and what they can infer. Measures include intelligibility, inference accuracy, and the user's own discomfort, often via audibility tests and privacy scales compared against a visual-only condition. Note that real leakage comes from both sides of the conversation, not only the device's own cue.

Where it stops holding

When content is not sensitive—"battery low"—risk is low, and extra silence may cost usability for little gain. Private channels such as headphones and bone conduction cut leakage sharply, though physical contact and visible devices still reveal something. Tolerance for a given level also varies culturally, so the same loudness is judged differently across settings.

Applying it

• Route personalized content to headphones or a private channel by default; keep speakers for non-sensitive generic cues.
• Grade content: short public-safe cues, specific content requiring a private channel, and sensitive content that must not sound.
• Show a real-time privacy cue whenever output goes to a speaker, so users know how much is audible.
• Verification: record speaker output at typical noise and distance, then have uninvolved listeners restate what they heard and what they inferred; use that to decide whether a private channel is required.

Related

• Within the group: D2.08.2 How recognizable a cue is becomes a social label · D2.08.3 Sounds on by default need a prominent off switch
• Adjacent: D2.11 Muting and category control for sound · D2.05.1 Critical information must move to another channel when sound is muted
• Search terms: acoustic leakage · overhearing · privacy in public