There is a mistake in updating beliefs around norms that I have seen multiple times in different contexts made by different people. This is my attempt to classify this (fallacy? /mistake?), in order to give it a name and make it easier to refer to.
A friend of mine was recently honestly surprised that two government politicians got into trouble for personally taking money from mask-producing companies in return for securing government orders. He was not surprised that the politicians were corrupt. He was surprised that there were any repercussions. He thought that politicians could totally get away with this. How did he acquire that belief? A while earlier it was revealed that another politician received stock options from an IT startup, for which he then put in a good word at the government. He only got shit from the press, but his career is still going well.
My friend had previously thought that there was a strong norm against politicians being corrupt and that any violation of this norm would end their career. The politician who got away with the stock option affair was strong evidence against such a norm being enforced. Thus, he updated towards the belief that corrupt politicians can just get away with anything.
But how does this constitute a violation of rationality? Wasn’t I saying that this was some sort of fallacy? What did he do wrong?
His mistake was this: He assumed that, if there was a norm against corruption, it would follow a definition of corruption that did in some way actually corresponds to the degree to which the act was harmful. Let’s look at this in a graphical way:
Norms here are represented with lines, separating the forbidden acts on the left of the line from the allowed acts on the right of the line. His hypothesis space only consisted of straight lines. His prior was that there was a norm enforced (norm A). Then he observed act 1 occur unpunished. This made him update that there was either no norm or maybe a norm as far out as norm B. Then he observed act 2 being punished and was confused. The true norm that was enforced all along was norm C, which looks pretty gerrymandered. Does it make sense that MPs are obliged to formally record monetary payments they receive, but not stock options? Not really, but that is how it is.
I have also seen the same play out in other contexts. For example, after the FSB (probably) killed a former Chechen War commander in Berlin, my roommate thought that the Russian secret service would just kill anyone they held a grudge against and expected a journalist might be their next target after making fun of Putin. I think that expectation was totally unrealistic. It stems from the assumption that all lives should be protected equally. But in reality, if Putin kills a rebellious war commander or a double agent, other countries react with consequences that he is willing to take. But if he started killing western journalists, NATO would probably get really mad.
I think as a rule of thumb, we should expect gerrymandered lines in the sand when powerful people try to get around some norm.
So, next time, you see some powerful person get away with something you didn’t expect them to, consider that there is a line in the sand, just not where you think it is.
Modeling behavior as if there are specific lines might not be the best way. Your friend isn't the only one who thought that the politicians who received the money in the mask deals could have gotten away with it. Quite obviously those politicians also thought so at the time they made the deals. Those politicians in turn likely do understand the environment in which they are operating and how the lines are gerrymandered.
Different people have different political strengths and can defend themselves in different ways when they are at the center of a scandal. The metoo allegations against Joe Biden did nothing, not because there aren't norms, but because it wasn't in the interest of the Democratic establishment to enforce norms.
I expect that most of the involved politicians would say that the mask deals were more harmful than the letter that Amthor wrote. I do think they believed that the COVID-crisis is very serious and that acting badly in it's context is very harmful. If you want to get people to follow a mask mandate, having politicians profit from deals with mask manufacturers is serious harm.
That's a mechanism about how disclosure works but not one about the political pressure to resign after the scandal came public. I also don't think that the way the disclosure rules work here is a feature of gerrymandering. It's just that the German disclosure rules are not written with MPs being employed by startups that pay their board in stock options in mind given that most MPs work in other settings.
This summary ignores what the case was about. It's legal for a German MP to put in a good word with the government. An MP is free to be hired. The thing that's not legal for a German MP is to put in a good word for someone that pays them in their role as MP. Amthor used the letterhead of his parliamentary office and that's what makes it illegal corruption. If he would have actually used the letterhead of being a board member of the corporation, that letter would have been legal.
When it comes to the mask deals then the allegation is that the relevant MPs acted in the scope of their job as MPs when they arranged the deal and it wasn't just a matter of using the wrong letterhead to make it legal. The mask deal is also hard to imagine without someone explicitly telling the MP that they are getting the money for making the deal.
With Amthor on the other hand it was not clear that anybody explicitly told him to do what he did with the government. And that does matter legally.
I think two mistakes in your friend's model. The first is simple over-correction - seeing one instance and believing that's universal. The second is over-simplification, which is what you're pointing at with this post. People are complex, and most social decisions are heavily context-dependent. Some people get away with things that others don't. the very concept of "norm" is named for "normal", and is about the median/center of a set of behaviors. Forgetting that people are actually on many distributions, which can have pretty long tails, is the error.
Suggestion for an alternative model, simpler than gerrymandering: prob(punishment) depends positively on the severity of a norm violation, but there is no threshold where it becomes 1 or 0. Even though you draw a two-dimensional diagram, You model seems to have only one dimension, and so there is some randomness capturing the things left out of it.