Hacker News Reader

Comment by jameslars | Hacker News Reader

Did you look at the original? I assumed it was censoring "killer" which also meshes with my perception that "trigger words" get censored in certain social media circles now.

View on HN

Most of this is self-censorship by people who believe they are being oppressed by all-powerful dark forces, whose malevolent schemes can nevertheless be foiled by simple character substitutions.

My favorite example of this is people self-censoring “shooter” as “sh**ter”, which I can’t help myself from always parsing as “shitter”.

“The shitter in Virginia has been apprehended”, etc.

To be fair, YouTube (and probably other platforms) used to literally block stuff like that, and the substitutions did get by it.

Facebook does this heavily. People discussing the strategy to beat a Fallout boss get sent to Facebook jail for writing "gun", "shot","kill", etc., so the self-censoring is now in the rules of most Fallout gaming groups.

I assume it's just people mimicking/cargo culting content creators, who are indeed being oppressed by the dark forces of the algorithm and the de-mometizer.

> Most of this is self-censorship by people who believe they are being oppressed by all-powerful dark forces,

What a ridiculous take. Many platforms and games actually prevent you from posting certain words, or flag or otherwise penalize posts if you include them. In some cases, this is completely explicit, i.e. you'll get a message saying "cant' post that."

Instead of playing a guessing game of which platform censors what, it's common just to avoid such words by censoring, euphemisms, or synonyms.

How did "kosher" become one of those words?

IMO this boils down to big platforms being pressured to moderate, because some horrible things are shared there, and the cheapest moderation is blindly flagging words. People discussing things like games get warned/banned/shadowbanned by machines for using banned words, and everyone is getting used to self-censor because the appeal is as useful as a second navel.

TL;DR: self-censorship is so prevalent because it saves online platforms money.

ok, so i misread the censored word. Maybe i'm old; the concept of "trigger words" doesn't seem to be solved by making it harder to communicate in my opinion.

Definitely agree. I don't understand how "thinking harder" about what the word MIGHT be solves any problem. I might just be part of this older generation who is hard trained on reading and not skimming, I don't know. I've never found anyone who appreciates this sort of censorship to ask them to help me understand.

Nobody appreciates it. It's in the Nash Gap -- the difference between the true optimum and the reality that emerges from each involved party optimizing for their incentives.

People: complain to politicians when egregious content starts trending

Politicians: bug Platform to do something about the egregious content

Platform: quickest fix for egregious content is dumb pattern matching

Content Producers: avoid becoming a false positive by using dumb substitutions to dodge the dumb pattern matching

The most direct fix would be the platform getting smarter about censorship, but this would not be without its own tradeoffs.

It not /all/ about triggers it's partially driven by the algorithms on some social sites like TikTok or Youtube where there's some evidence and a lot of speculation that mentioning or including certain words can cause the recommendation algorithm to suppress your videos etc. That leads to a lot of homophones etc popping up like "sewer slide" or odder non mentions like unalive popping up. Because the really big channels care about maximizing reach for monetization reasons it shows up strongly in peoples feeds which drives wide adoption.

Beenaminute did an experiment about this and found it didn’t really matter. Algospeek would be caught by the same filters people try to avoid anyways.

There'd be a lag between a new algospeak abomination popping up and it getting added to the filters at least but yeah I think it's probably largely ineffective now. Probably even less so if they're using LLMs to screen for potential new substitutions but the practice already gained popularity so even if it's ineffective now it'll persist until it's supplanted or becomes outdated. It's like the persistence of some SEO optimization techniques even after Google et al adjusted the algorithm to account for them.

Reply on news.ycombinator.com