# The Loose Wire — case notes

*Research file. The story lives in [the storyboard](comic-loose-wire.html) and the radio
script — this holds only the grounding neither surface can carry. Nothing here restates the plot or
references a panel number, so nothing here goes stale when a draft is rewritten.*

**Lead:** Connie Chase. Adherence and the roster are coordinator work.

**Feature:** wear-time monitoring. The platform computes what fraction of each day a device was
worn and raises a participant whose daily figure stops moving. Note what it is **not**: it does not
record when the device came off, and it never says why. A percentage and a count is the whole of
what it knows.

**Trial:** the **VIGIL** trial. A wearable study where the most compliant participant looks like the
least.

---

## What the case rests on

**The culprit is a threshold and an instruction card.** Wear time says how long a device was off.
It cannot say why. A cutoff decides who stays in the analysis; nothing anywhere decides whether the
missing hours were indifference or obedience, and the study had already assumed the first.

**A summary conceals a series.** The reported figure is correct and hides everything interesting.
The clue is in the daily numbers underneath it, which never vary, and which appear on no report
anybody prints. This is the methodological heart of the case and it costs the product nothing to
make: a wear-time percentage per day is already computed, and no clock time is ever needed.

> The Quiet Participant was *absence doesn't print*.
> The Invisible Line was *staleness doesn't print*.
> The Perfect Delivery is *sent was true, delivered was true, and both were useless*.
> The Loose Wire is *the average was right and told you nothing*.

### Nothing here is a sensor failure

This is the part most likely to get drafted wrong. The device correctly recorded that it was not
being worn. The report correctly totalled it. The PI correctly quoted the total. **Every instrument
worked.** The only broken thing is the inference laid on top of them, which is why the fix is a
sentence on a card and not a better watch.

Specifically: do **not** argue that a wearable cannot distinguish a still body from a still object.
Modern devices detect wear directly, and the claim is false.

### Why a human could not have caught it

The failure mode is not inattention, it is explanation. The gap was noticed early and closed with a
theory that fit — *she's easing out, they all do* — and acted on weeks later. **The danger in a
study is almost never the data nobody looked at. It is the data somebody looked at once and
explained.**

The correcting evidence is a distribution, not an observation. A person carries an impression;
only something that counts carries a variance.

---

## What the case argues

The instrument did not diagnose anything and could not have. It noticed that a number had stopped
behaving like a person, which is a shape rather than a cause. The answer came from asking the
participant.

That is a more honest claim than "our system catches non-adherence," and it is the one that's true.

**The assumption under it all is missing-at-random.** Treating a gap as noise is the default every
analysis quietly makes, and this case is the ordinary violation: the hours are absent for a
systematic reason, and the reason is the study's own paperwork.

---

## Notes

**Nobody is a fool.** The participant followed her instruction card exactly. The PI quoted a
threshold she wrote into a plan for good reasons, and it is a good threshold — being inconvenienced
by your own rigor is the correct way for her to be wrong. The coordinator did the job the protocol
asked with the instrument the protocol gave her. The card is the only thing at fault, and a card has
no feelings.

**The kindness is the trap.** The advice to let it go should come from someone being decent at the
end of a long week, not from someone incurious. A study is far more likely to lose a participant to
a colleague's generosity than to anyone's negligence.

**Don't let the participant become a lesson.** She is a person who did what she was told and had no
way to know it was the wrong time to do it. She is never scolded, never corrected, and never
explained to the reader.

**Register:** the wiretap. Informants, the wire, a source going cold, an alibi, money slang.

**Participant 031**, as always, and nobody remarks on it.

---

## Open

- What VIGIL stands for. The backronym is unwritten and the radio version will want the whole joke.
- Whether the participant keeps a name (Mrs. Ruiz in the current comic draft) or stays a number.
- Whether the lost nights are stated out loud. The comic currently doesn't, which reads as a clean
  save; the fiction says weeks of nights are gone for good. The radio version has room to be honest
  about it and probably should be.
- Three Connie cases out of the first five. Correct by the assignment rule, worth watching.