Passive collection differs from active query, and usually goes unnoticed

What it is

Actively querying data (opening the app to ask "what's the temperature at home") and passive collection (sensors recording continuously, waiting for no one) are two different data relationships. Active querying carries its own awareness — whoever initiated the action knows they pulled data. Passive collection has no initiation moment: the collecting does not coincide with any user action, so there is never a moment at which "I am being recorded right now" becomes salient. Ambient computing's normal mode is precisely the latter.

"Usually unnoticed" is a description, not an accusation: smaller devices, screen-less devices, sensors embedded in furniture and luminaires — the design direction actively lowers the perceivability of "there is a sensor here". Awareness needs cues, and eliminating cues is the success criterion of this style of design.

Why it happens

Three layers to the systematic unawareness:

• No behavioural anchor. Awareness usually attaches to "what I did" — flipping a switch, opening an app. Passive collection is asynchronous with every user action and lacks the action that would direct attention to the data flow.
• A persistent state cannot be perceived in any single moment. "Collecting all the time" is not instantiated in any single instant; people notice events, not baselines. Yet the risk and the profiling accumulate precisely in the baseline: each individual stream (temperature, presence, door contacts) looks harmless; the aggregation is the complete picture of a household's rhythm — and aggregation happens backstage, where the quantitative drift never enters anyone's experience.
• Physical presence is designed away. The sensor's physical body is the last cue, but consumer-electronics aesthetics (small, hidden, blended into the environment) run directly against the privacy requirement of "discoverably present".

So the affected party's picture of the household's data flows rests on installation-day memory and stale common sense — while devices multiply, firmware changes, and aggregation logic evolves, the picture stays frozen on day one.

Studying it

• Interview studies: work on informed consent for ubiquitous data collection (e.g. Luger and Rodden's interviews on consent in context-aware computing) found users struggle to grasp the scope of continuous collection and lack any real-time sense of "being collected". The SOUPS 2017 smart-home interviews (Zeng, Mare and Roesner) likewise document gaps between what devices collect and where it goes, and what users believe.
• Awareness-rate measures: on home visits, ask members to list "the sensors currently collecting in this home, what each collects, how often" and compare with the actual deployment for a recall baseline — a measure that reflects the present state better than "did they ever consent".
• Experience sampling: probe in daily life — "right now, what do you think is recording in this home?" — and compare against ground-truth collection state to measure the direction of the gap.

One methodological caution: the probe itself creates awareness (asking is a cue). Measure the baseline unprompted and in a natural state before introducing any intervention.

Where it stops holding

• Unnoticed does not mean unconcerned. Once explicitly told "this device records your daily rhythm", most people react strongly — the risk exists, it simply was not perceived; "nobody complains" cannot be read as "nobody minds".
• Devices with screens and apps get noticed more. Anything with a visible interaction surface (speakers, screen hubs) is remembered as "an electronic thing"; the dangerous cases are the fully faceless ones (sensing modules hidden in luminaires, sockets, smoke detectors).
• Long cohabitation dulls awareness. The vigilance of a new device's first two weeks erodes into habituation — "knowing it's there" decays into background noise. Awareness is not a settled state measured once; it decays.

Applying it

• Preserve discoverability of presence: keep one factual entry point per sensing device — an inscription on the device itself ("contains presence sensing and a microphone") or the front page of the packaging; faceless devices need it most.
• A standing inventory, not one-shot notice: a persistent overview of the household's data flows (which devices, collecting what, stored where) in the app or a local page. Seeing it once at installation is not enough — devices and aggregation logic change, and the inventory must change with them.
• Devices with sensitive capabilities (audio/video, continuous physiological signals) should be physically identifiable: form, indicator, or placement that lets a person point at "that thing is collecting".
• How to check: unannounced home visits measuring unprompted recall of "the sensors collecting in this home", re-measured after deploying the inventory entry point. The gain in recall is the measure of the notice design.

Related

• Same group: Z6.06.2 After-the-fact notice cannot replace a prior opportunity to know · Z6.06.3 Notice must match the situation; dense terms do not work · Z6.06.4 Short-term guests and long-term residents need different depths of notice
• Nearby: Z1.05 The costs of invisible computing · Z6.08 Physical indicators of collection
• Search terms: passive sensing · ambient data collection · awareness of collection · smart home privacy