Camera coverage routinely exceeds what the owner expects

What it is

The purchase decision for a home camera rests on a target scene — the front door, the cot, the pet. But physical coverage is not bounded by purchase intent: field of view, pan-tilt movement, and microphone pickup make the actual footprint far larger than the picture in the owner's head. The owner thinks "the hallway"; a 130° wide lens takes in the whole living room, sofa and dinner table included.

Two groups are surprised. The owner routinely underestimates coverage; everyone else in the household — and guests — can only assume the worst. Smart-home research calls the latter secondary users: they never bought the device, never installed its app, yet stand inside its field of view.

Why it happens

Three sources of systematic deviation:

• Optics does not follow semantics. The owner's mental frame is semantic ("watch the door"); the camera's coverage is geometric (wide angle plus depth of field). Semantic intent has no grip on physical parameters — a wide lens that "sees the whole doorway" necessarily admits what lies either side.
• Coverage is volumetric and mobile. Pan-tilt cameras can be steered remotely, changing coverage in real time; nudging the device or rearranging furniture silently shifts what is in frame — while the household's shared understanding stays frozen at installation time.
• The microphone is wider than the lens. Audio pickup extends beyond the frame boundary; conversations happening outside the picture are still captured. What is visually "not covered" may be acoustically covered.

The structural mismatch: privacy boundaries are drawn by rooms and relationships (the bedroom is off-limits, the living room is fine), while capture boundaries are drawn by optics and acoustics. The two boundary sets do not coincide; where they do is luck.

Studying it

• Interview studies: the SOUPS 2017 interviews by Zeng, Mare and Roesner on smart-home security and privacy concerns are a standard anchor — owners were most anxious about cameras and microphones, explicitly recognised that family members, guests, and nannies would be covered, yet almost never configured anything on their behalf.
• Expectation-versus-measurement audits: have the owner sketch "what you think it sees", then overlay the measured field (live view frustum, taped boundaries on the floor) and quantify the direction and area of misestimation.
• Longitudinal follow-ups: track how coverage drifts with furniture changes and device relocation, and how long the "installation-time consensus" actually lasts.

One methodological caution: measure expectation errors in real installations. A lab mock-up lacks the private objects and circulation paths that "happen to be in frame" at home, so the error is systematically underestimated.

Where it stops holding

• The error is asymmetric. Underestimation (thinking it captures less than it does) dominates, being guaranteed by the physics of wide angles and audio pickup; overestimation mostly appears with night vision and backlighting (installed, but unusable).
• Outdoor doorbell cameras involve different subjects. Their expectation gaps chiefly affect passers-by and visitors rather than cohabitants — a public-space notice problem with its own dynamics.
• Physical masking keeps the gap manageable. A lens shutter or software privacy zones (detect and record only inside a drawn region) can pull coverage back to the intended scope — provided users know the features exist and use them.

Applying it

• Calibrate at installation: force a live-view walkthrough during setup, letting the owner pin the field boundary on screen or mark it on the floor — "here is what you will capture" becomes visible, and expectation errors surface on the spot.
• Offer privacy masking by default for sensitive orientations: regions of the frame (the sofa, the changing route) that are never detected or recorded, rather than an all-or-nothing choice.
• Re-confirm on movement: a pan-tilt returning to a preset, a camera moved to another socket — both are coverage changes and should re-trigger the framing confirmation.
• How to check: on a home visit, have the owner and one non-installing member each sketch the coverage and compare against the measured frustum. Both groups' errors converging inside the masked zones is the pass condition.

Related

• Same group: Z6.05.2 Recording and live viewing differ in privacy risk level · Z6.05.3 Family members diverge sharply in camera acceptance · Z6.05.4 Storage location determines the actual risk of exposure
• Nearby: Z6.03 Privacy in shared spaces · Z6.06 Notice for passive collection
• Search terms: field of view · secondary users · smart home camera privacy · privacy masking