What you reveal — invisible layer

experiment · motion inference

your movement
is a confession.

Apps don't need a camera to know what you're doing. The way you walk, type, drive, or sleep — all of it shows up in the accelerometer. Walk around for a few seconds and watch what can be inferred in real time, without any special permission or any awareness from you.

your typing rhythm is as unique as a fingerprint. Research at MIT and Stanford showed that accelerometer data — collected silently by apps in the background — can identify you with 99% accuracy from how you hold your phone while typing. The same data predicts your age, gender, health conditions, emotional state, and whether you've been drinking — at the level of individual drinks consumed.

—activity

—step Hz

—variance

press start and move around to see inferences build up

▶ what can really be inferred from motion?

Academic research (published, peer-reviewed) has shown that phone accelerometers can reveal:

Activity state — sitting, walking, running, driving, cycling, sleeping
Typing content — keystrokes inferred from micro-vibrations (MIT 2011, 99.5%)
Gait fingerprint — individual walking pattern is unique (95%+ across 100m)
Age & gender — stride length, cadence, vertical bounce vary predictably
Health conditions — Parkinson's gait, fall risk, COPD exercise capacity
Emotional state — higher variability correlates with anxiety; low variance = depressed
Intoxication level — BAC 0.08+ detectable from gait instability (80% accuracy)
Floor number — step counting + barometer = which floor of which building
PIN entry — tilt angle while typing reveals which side of screen you touch

None of this requires any permission. Motion sensors are freely accessible to every app.