Instrumented People, Places and Things: office beacons, 2014

As a way of seeing where a few people went and how long they stayed, yes. A walk reads as a walk, a desk as a desk, and a day as long stays joined by short trips, with no camera and nothing to wire. The picture is at the scale of an aisle or a group of desks, and not of a chair.

Its costs are also visible. It knows only about people who wear the iPod and leave it running, and the team's tracker has wearers plugging themselves in when batteries died. Beacons were put up by hand and their places marked on the app's plan. Near edges and in the elevator core the position fails or is pulled inward. The second phase was simpler for the wearer, but seven listeners give a place in a room rather than a path through it.

A beacon only announces itself. The iPod listens and logs each beacon's number and signal strength (RSSI, in dB). The team's viewer then did this:

1. A scan is all readings less than 125 ms apart. An iPod made about one a second and heard about a dozen beacons in each.
2. Readings at or below −90 are dropped. Each of the others gets a weight: its signal mapped from −56…−100 to 50…5, to the sixth power, so the loudest beacons decide.
3. The position is the weighted mean of those beacons' recorded places. No distance is computed, and a position can never fall outside the beacons around it.
4. The path is averaged over seven scans.

Run on the surviving logs, these steps give the team's own computed positions: 281,484 of 281,485 scans are the same.

Added here: the floor. The team's viewer read one floor's beacons, so a device upstairs had no position there. This page lets each scan vote for the floor whose beacons are louder, takes the majority of 21 scans, and places the scan among that floor's beacons.

Path

The team's drawing: a small ring a scan, fading over 240 seconds, and a heavier ring at the head, wider when the iPod's microphone was louder. The readout at the top left is theirs too.

Beacons

Also the team's, from their June viewer: blue circles are beacons heard in this scan (on the Day view, this minute), wider when louder, outlined when above the threshold; gold dots were not heard. The thin lines to the head, drawn when paused, are added here. On the walk of 11 July five beacons still lay near the desk, far from the places recorded for them later; they appear in the lanes and not on the map.

Lanes

For every device, a row under its label: its motion in its own color, as the team drew it (the iPod wrote a row only when it moved), or, on days that kept no motion, a pale band while it logged, colored where it read as upstairs. Follow one device and its lanes open up: motion in three axes, sound level, and Heard, a row a beacon, ordered along the floor, gray below the threshold and blue above it, darker when louder. Click to move the time.

Dwell

The team's heat pixel, 2 ft across, and its four colors. The team's colors counted scans, one to four; the limits of time here are added, and the legend gives them. An iPod set down on a desk counts the same as a person sitting.

Labels

Each track is shown under a label, with no names. A label follows the iPod, and two iPods changed hands between July and August, so a label isn't always one person. The second phase's labels are letters, and nothing joins them to the first phase's numbers. Click a label, or a head, to follow one device.

Within 15 ft

Added: the dashed lines on the plan and the Together view. Two iPods within 15 ft on one floor, for a minute or more. Most of it is neighboring desks. Away means both iPods more than 20 ft from their own commonest place that day.

Floors

Added. A row is one iPod on one day, under its label.

Walk

No record of who walked or by what route survives. That it is a lap is our reading of its shape.

Scale

Feet are from the office's plan.

Volunteers. Each collected an iPod in the morning, wore it in a sport armband and returned it before leaving. The team's consent form and the app's own consent screen survive; signed copies were not found. The form lists what was logged: motion, sound level, obscured images, signal strengths and surveys.

Each track is shown under a label, with no names: TJ's own as TJ's, and the others as numbers and letters. No device identifier and no name of a person is published, and nothing here says which wearer was on the team. iPods changed hands, so a label is not always one person.

At least four of the seven iPods were worn by the Iota team on each surviving day (counted by hand, matching the team's tracker to its list of members). So these are real working days, but they are mostly the team watching itself, and not the office going about its day.

The sound level is a loudness number. The recording it came from was thrown away on the device.

Survives: raw logs for four working days, 31 July and 7, 8 and 11 August 2014, with both floors; the team's computed positions for 12 to 15 August, for the lower floor only; shorter logs from one device, 28 May to 15 July; both floors' plans and beacon places; the team's viewers and app. Of the second phase, one hour from all seven iPods (23 October) and one iPod's log of 2 October.

Not found: the rest of June to September, known only from lists of what was uploaded on some days; the rest of the second phase, which ran to early November with thirteen beacons, ten of them heard in the surviving hour; the surveys; any analysis of dwell or of people together. The drive's index was searched by name and date. It stops at July 2020.

Kept off this site: the iPods' low-resolution test pictures of the office, which show people, and a picture the team saved from its viewer, which shows the devices' identifiers.

It could tell which part of which floor a device was in, when it moved, and how long it stayed. It could not tell what anyone was doing there, or whether an iPod was on an arm or on a desk.

Collaboration and movement across two floors were hoped for. Neither was measured by the team in anything that survives. The logs do show who was near whom: on six of the eight days, some pair spent ten minutes or more within 15 ft of each other and away from their usual places. Whether they were working together, the logs can't say. Walking together barely shows: at about a reading a second, two iPods moving side by side for more than a few seconds is rare in these logs. Movement between floors is there, for the few who made it.

TJ writes: “An observation is not an activity model. A sensor may report that a door opened, a cup moved, or someone entered a room, but none of those events alone establishes what the person was doing.” (Modeling Activity, p. 3)

And: “Models remain selective. A sensor has a range and resolution; an event model establishes boundaries; a spatial representation omits aspects of experience.” (Computational Design, p. 4)