Instrumented People, Places and Things: measuring collaboration, 2013–14

Method

The team's sensor matrix of 3 May 2013, row for row. Click a factor, behavior or piece to follow it; the sensors' counts are made here from the marks and agree with the sheet's own.

Office

The plan is redrawn from the report's, p. 20. Spiders were Iota's own sensors, and Grid-EYEs thermal cameras of 8 × 8 pixels. The rings at the front door are spider 7005's seven bands at the replay's time, wider and warmer louder, as on the open house page. Dashed instruments have no readings to play.

The day

The lanes in color are the surviving log. Its sound is a spectrogram: the seven bands as rows, 63 Hz at the bottom, each band's highest reading in each stretch of time, on one log scale for all seven; point at it to read one. Above it, in black, their magnitude. The spectrogram is TJ's, from 2026. Gray lanes are traced from the report's figure (p. 39), for shape only. Scroll on the lanes to zoom, drag to pan, click to set the time. Closer than 25 minutes they draw all eight readings a second. Temperature and air pressure were what the report calls over instrumenting, which “can save a failed dataset or shed light on a strange pattern” (p. 36). The last lane adds the 2014 platform's 15-second step to this day's sound, which the team did not do.

Gray lanes

Traced for shape, to about 20 seconds. The clusters are the report's sorting of the meeting space's sound, a row to a cluster, in two runs: 2 to 3 pm and 3 to 5 pm (pp. 44–45). The entrances and exits are read from its p. 42, to about a minute.

Platform

The 2014 deck's own figures, which open larger when clicked, and the demo's model, which you can try. Nothing there is a reading from this day.

Students

Their model, computed here by their formula, and their timeline traced from their page.

Survives: spider 7005's log for 16 May 2013: 687,167 rows, of which 32 were shifted by serial errors and are left out. Sound in seven bands, temperature and air pressure. A script from June 2013 cut a stretch from this same file, row for row, and its sound lines up with the report's front door lane to within half a minute.

Also survive: the team's sensor matrix, deployment plan and reports; its analysis scripts; recorder programs from 2014; the demo program (the source of the meeting models you can try here) and a 58-second recording of it running; and a mock client report of July 2014, for a made-up company, which is not drawn from here.

Not found: the other 8 instruments, the other 43 days, the Kinect's validation photos beyond the report's pages, the buttons' and survey's raw answers, and every recording from the 2014 platform's meetings. The drive's index was searched by name, date and instrument number. It stops at July 2020.

From the report, of the 44 days: “It is easy to see when the corner meeting space is occupied by a meeting, if people are in the kitchen, or if there is activity in the entrance area.” But “getting beyond this simple level to a real understanding about workplace performance becomes much more difficult.” Of 50 people offered activity logging, 5 opted in. The buttons drew 74 presses, about 75% better than expected and 8% negative: their place “increased the number of responses, but decreased their honesty”. The survey drew 9 entries.

The 2014 deck names four limits. Heavy phone or video conferencing: part of the meeting is not in the room. Unstable wifi. Organizations with privacy or security concerns: the camera sees faces, and the service must be prepared to limit or remove instruments. No calendar data or room audits: a rough count and no exact one.

TJ writes: “Models remain selective. A sensor has a range and resolution; an event model establishes boundaries; a spatial representation omits aspects of experience. More detail does not automatically address what has been left outside the model.” (Computational Design, p. 4)

And: “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)