Office socialization

These slides run from what the study asked
to what kind of picture its approach gives.

The study's anchor is a place,
and a container of candy is how it looked.

“So the place is the office, the activity is socialization, I think.”

For eight weeks in 2012, 24 open containers of candy stood across a Chicago company's offices, from busy walkways to small kitchenettes. Each stood on an instrument that stayed put while the office moved around it.

Quotation: TJ, 31 October 2026, with its typos corrected. People, places and things are the study's anchors, and activity is what they share.

The question was whether cheap instruments
could reach an activity through its trace.

A proof of concept: any number here describes these few devices and no others.

Quotation: the Iota team, 2013. The scope of the work, as TJ has described it, “did not include generating findings or interpreting the data”.

Most of the readings are gone,
and fragments survive inside the team's tools.

The team's plots of the rest survive, and anything traced from them here is gray.

The eight-a-second records of the 24 instruments have not been found.

Each container stood on a kitchen scale,
pulled apart and refitted.

A load cell, a temperature and air pressure sensor, and a microphone behind a seven-band chip, which kept levels and never sound.

Eight ticks a second. It wrote a row when something changed, while the microphone was on, or every ten seconds.

Fig. 1 · The shell and the scale

Drawing: the Iota team, with the container left out. The parts: the instrument's firmware, 2012.

The microphone heard about 2¼ hours
of a workday, in short windows.

So most taking has no sound beside it, and a voice is only ever a level across seven bands. Nothing it kept could be heard as speech.

Source: the instrument's firmware, 21 September 2012.

The team looked for events two ways:
by knowing what to look for, and by not knowing.

Method: the Iota team, 2012 to 2013. Sherlock was built in Processing and Java; its source survives. The shingling scripts have not been found.

A fingerprint is three numbers,
and how much each of them matters.

Source: the tool's own settings file. A count is one part in 1,023 of the scale's range, and rises as weight leaves. Energy adds each step from one reading to the next.

Seven takings in two minutes could be
one person or five: the weight can't say.

Computed here from device 1315's readings, by the team's arithmetic and its final fingerprints.

The model counts a trace,
and the question stays open behind it.

“An observation is not an activity model.”

Such a record, TJ writes, “could be evidence relevant to a social hypothesis without itself being a measure of sociability. The model's value depended on whether the relationship … could be examined critically, not on whether a dispenser could count events accurately.”

TJ, Modeling Activity, 2026, pp. 3 and 4.

What it could not tell was who,
how much, and whether anyone was together.

• Who, and with whom. The weight has no person in it, and the microphone no words.
• How much. “drift of the scale measurement (not being able to tare)”.
• Whether it was right. “acquiring data to validate events, capturing video (privacy concerns), and incomplete reporting on device locations.”
• Whether clusters mean anything. “Although it looks convincing, we don't know if it worked.”

Quotations: the Iota team, 2013. A container's cleaners, movers and refills are all in the trace too.

As a proof of concept the approach held,
and gave a picture of rhythms and places.

• A kitchen scale could see a hand go into a container, and a tool could sort what it saw into kinds that can be examined case by case.
• The office's rhythms came through: the work week, the start of the day, lunch, and the end of the day.
• Places differed: some loud and some nearly silent, with taking in all of them.
• Sound and taking were set side by side, in short windows, and read by eye.

These describe a few containers in one company's offices in 2012, and none measures offices in general.

Everything on these slides traces
to a small set of sources.

• Device 1315's weight: two exports in Sherlock's demo, December 2012.
• Device 1414's sound moments and the training set: the team's sound tools, November 2012.
• Sherlock, its source and a recording of it running; the fingerprints in its settings.
• The instrument's firmware, September 2012.
• The team's plots of three devices, October 2012, and its charts, 2013, traced.
• TJ, Modeling Activity, 2026, pp. 3 and 4. TJ's earlier site, on the scope of the work.

The study, the instruments and the tools were the Iota team's. Made in 2026 at TJ's direction, with Claude Code.