Artificially Intelligent

Any mimicry distinguishable from the original is insufficiently advanced.

Contextualized Norms

| 963 words

There are certain questions you’re not supposed to ask; “do disabled people have less moral weight?”, “does gender correlate with intelligence?”, or “are certain races more prone to violence?” are some examples. People that lack social savviness are really confused by why they aren’t allowed to ask these questions. People that have social savviness might say something like “because it’s rude” (I don’t claim enough social savviness to be able to predict properly). The concept of a norm being contextualized provides a generalized answer to why you can’t ask certain questions.

Contextualized norms are, as the name suggests, sensitive to context. When you ask a sensitive question, that question has context. Most of the time, the context of that question is a history of discussions designed to delegimitize certain persons. A more complete answer to why you can’t ask some question about X:

There are some topics that aren’t quite open for discussion. Historically, nearly all discussions around X have been initiated as an attempt to cause harm to certain people. To counteract this, as a society we have decided that talking about X is no longer permitted except in certain spaces, after certain caveats. When you ask a question about X, you probably don’t want to hurt anyone. However, given the historical context, there’s no way for you to signal that you want to have a genuine discussion.

Compounding this problem is the fact that historically, people have said “I just want to have a genuine discussion” as a way to start disingenuous discussions to cause harm. In general, if there’s any cheap signal that you can send that allows for X to be discussed appropriate precautions, malicious actors will learn to mimic that signal. Thus, the norm that certain parts of society have settled on is that X is just not allowed to be discussed in certain places. Given that this norm has been established, the only people that generally try to start discussion about X are malicious actors. Thus, whenever you ask questions about X without first robustly establishing that you’re not a malicious actor, you are assumed to be a malicious actor.

Contextualized norms require that you include caveats. They insist that you first acknowledge the relevant and necessary historical facts, point out mistakes society has made in discussing similar things in the past, recognize that you probably come from a place of privilege, etc. Contextualized norms require that you send an expensive signal that you are the type of person that can discuss X without causing harm.

(The phrase “check your privilege” might be able to be interpreted as “you haven’t included the necessary caveats to make me confident you’re not a malicious actor, so I’m going to assume you are until shown otherwise”.)

In general, contextualized norms amount to taking the outside view on discussion topics - they say something like “if we look at all the times this thing has been discussed without extreme caution, it’s reliably caused suffering; therefore, we’re not going to talk about it”. These norms often are just blanket bans on certain topics to make them robust to malicious actors.

Contextualized norms don’t always apply. If you’ve interacted with a person for long enough and they’ve properly navigated contextual norms enough times, then you can infer that they’re probably not a malicious person. You can thus discuss sensitive subjects without the necessary caveats because you know that they already know what the caveats would say.

But there’s more nuance (who would have guessed?). Suppose I’m friends with Alice and Bob. I’m close with Alice and have a “can violate contextualized norms” relationship with them. I’m friends with Bob, but not extremely close. Alice and Bob are acquaintances, but not close. If I’m with Alice and Bob together and Alice violates a contextualized norm, I will generally act as if Alice violated the norm and we didn’t have a “norm violations OK” type relationship. This can be very confusing for Alice. If I had to explain to Alice why I reacted differently because Bob was there, I would say something like:

Alice, we have the sort of relationship where we can talk about difficult topics without providing a bunch of caveats because I’ve spent a lot of time with you and I know you to reliably not be a malicious actor. However, Bob has not spent that much time with you. To Bob, you attempting to start a conversation about a difficult topic without providing the necessary context makes him think you’re a malicious actor. Since Bob doesn’t know that I know that you’re not a malicious actor, if I act as though I know you’re not a malicious actor, Bob might think that I am in general willing to talk about sensitive topics without necessary context. To Bob, this suggests that I am ignorant of broader norms established to prevent malicious and harmful discussion from taking place. In order to assure Bob that I don’t aid malicious actors, I have to say something like “Alice, don’t forget " when you attempt to talk about a sensitive topic without providing context.

The key is that while Bob and I both might know that Alice wasn’t malicious, I didn’t know if Bob knew that I knew, so I had to act defensively. If the previous statement didn’t make sense to you, this general concept is explained in more detail here.

Social situations are complicated. Usually, people do fine with navigating using their intuitions, but for people without strong social intuition, it can be really hard. When someone violates a contextualized norm, it’s easier to assume they’re malicious than to integrate evidence and actually check whether or not you think they’re likely to be malicious; I recommend you check anyway.