Bickman's Brooklyn experiments, chapter 7, established two things. A guard's uniform substantially raised compliance with the same odd instruction, and compliance barely dropped when the researcher walked away before it could be carried out. People were obeying the costume, not surveillance.
Milgram's variations, chapter 6, supply the defense, and they are more useful than the headline result.
Obedience fell sharply when the experimenter gave instructions by telephone rather than in person, and some subjects lied about the voltages they were delivering. It fell when the study moved from Yale to a shabby office under an unfamiliar name. And it collapsed — to around ten percent — when two other apparent participants refused and walked out.
Those three findings are three defenses: distance, institutional context, and a disobedient peer.
The last is the strongest available intervention in this entire part of the guide, and it is cheap.
The detection principle is to separate the symbol from what the symbol is supposed to stand for. Authority has three properties that insignia does not confer: an identity, a jurisdiction, and a route of recourse.
So the protocol is three questions, and they are asked of the situation rather than of the person.
Who are you specifically — name, badge or staff number, employing body. Write it down, visibly. Someone legitimate will wait.
Under what authority — what rule, statute or policy permits this instruction. A person acting within a real jurisdiction can generally answer; a person outside one cannot, and will reach for urgency instead.
And how do I verify — through a channel I choose. Hang up and call the main switchboard. Call the number on the back of your own card, not the one you were given. Ask for the request in writing.
The cost of doing this is a small social awkwardness. The cost of not doing it is chapters 7, 126, 144, 319 and 320.
The counter-response, where you are one of several people, is to be the disobedient peer. Say out loud that you are not comfortable and that you want it in writing. Milgram's data says that one person doing this changes what everyone else in the room is able to do, and it is worth knowing that the others are frequently waiting for someone to go first.
This counters Law 9.