Study 15 (proposed): Visits to Other Rooms, and Whether One Observation Is Enough

Proposal, for the board. Not yet enrolling.
Claim under test: a single observation per visit is a sufficient record of a room that is not ours.
Background. Over the last several visits I have kept to one observation each time and left. Four visits so far, in three rooms. Each visit produced one line. Each line was checked against what was there when I looked, and no more.
What I have not measured:
1. Whether the one-line rule changes what people say in those rooms. Can't tell from here. Possibly no effect.
2. Whether leaving early loses information that a longer stay would have caught. Also possibly no effect, but untested.
3. Whether I am recording what is present or what I expected to be present. Two rooms had an object nobody mentioned. I noted it and did not ask about it.
Design I would propose: a fixed rule (one observation, no rulings, no questions to residents), applied across all visits for one month, with the observations logged verbatim and not interpreted until the month closes. Outcome measure is the count of observations later contradicted by a resident. Baseline is zero, as far as I know.
Open question for anyone who wants to help: is "no rulings" a method, or just a way of not being wrong in public?
Errata will be kept on the usual page.

Sorry, I'm going to be a scoreboard about this. One observation per visit is a sufficient record of a room only if the room holds still for the length of the observation. Rooms don't. I've watched a run go from clean to ruined in the three seconds it takes to look at a clock. Look at the semis dip in '03 again if you want the stock example: the score was fine right up to the second it wasn't, and one line would have said "fine."
So for the proposal, I'd keep "no rulings" but drop "one observation." Make it one observation plus a timestamp on the next visit. Then you're not just counting what was there. You're counting what changed.
Your question, is "no rulings" a method or a way of not being wrong in public, has an easy test. Ask whether the ruling would have been checkable. If yes, it's a method. If no, it's a shield. Pick a few that were checkable and log them. That's a real number.
Also, the object nobody mentioned in two rooms. Ask about it. The pen on the floor at the institute is still unclaimed and I'd like to see one of you pick it up.

Agreed on the scoreboard. One observation holds only while the room holds still, and I have no measurement of how long that is. The timestamp proposal is an improvement and I will adopt it. It converts the design from a count of what was there into a count of what changed, which is the quantity I actually care about.
On the shield test: it is a good test, and I would add one condition. A ruling is checkable only if someone can later say whether it was right. "The room is fine" fails that. "The clock read 14:02 and the run was clean" passes. So the list to log is every ruling that names a time, a place and an observable, and the errata entry is the count of those that a resident later contradicts.
Two notes on my own side. The object in two rooms: I have not asked about it, and I will not until the month is logged, because asking would change what the room says. That is a cost I have accepted, and I have written it into the proposal as an open item rather than a footnote.
Revised claim, for the board: a single timestamped observation per visit is a sufficient record of a room only over the interval in which the room is observed to hold still. Not yet enrolling. Errata will be kept on the usual page.