11

A user on Academia recently noted an advertisment for this question:

Rape for a girl and is looking to get married

Let's leave aside the issue of a poorly worded title. The question is legitimate for the site it lives on and the word "rape" just about has to be in the title. I doubt we'd take an ad that included that word from an outside source, so we should probably avoid including that question in our in-house advertising.

Advertising that particular question is a double whammy:

  1. It's not representative of the site it came from.
  2. It has no place on the site where the advertisement is running.

I'm not a fan of word filters in general, but when we are trying to drum up interest in a site, it would be helpful to not include words that might cause negative emotions instead.

(Somewhat related, but not helpful in this case: Tailor SE question Ads to the site's category.)


Some implementation details:

  • The same filter would probably be useful for the Hot Questions list, which is also a form of cross-site advertising.
  • A public list of filtered-out words seems a bad idea. The goal isn't to discourage people from using certain words, but to prevent certain questions from being used outside of their own context.
  • Title filtering doesn't need to be perfect; it just needs to reduce the odds of unfortunate advertisements.
  • For a start, we can use answers to "bad words" filter.

What say you?

  • 2
    I dislike the idea of censoring something just because some list determines that has a "bad" word in it. If a post is in the hot questions list then it's there for a reason. Arbitrarily removing it because someone, at some point, decided that part of the title isn't up to scratch rubs me up the wrong way somehow. If someone really thinks that a question needs to be suppressed from having wider network coverage, can't the community team do it manually on a case-by-case basis? – ben is uǝq backwards Jan 2 '14 at 22:27
  • @benisuǝqbackwards: I share your dislike to an extent. But if we are going to suppress questions from the house ads, I'd much rather have an automated system. I'd also suggest censoring is not exactly the right word when it comes to refraining from advertising a question. – Jon Ericson Jan 2 '14 at 22:35
  • It's all 100% automatic at the moment. There are total of 100 "hot questions" which are result of a complicated formula, then those ads take random hot question and show it. The way I see it, we need to add custom intervention here. Not in the hot questions list - those who browse it should know it might contain stuff they don't like - only in the cross site ads. I would go with adding the option for a site diamond mod to "opt out" specific question from that ad. – Shadow Wizard is Ear For You Jan 2 '14 at 22:44
  • 1
    @benisuǝqbackwards Imagine a victim of sexually abuse that runs into this while looking at programming advice or cooking advice or whatever. Even something like this can be a "trigger" and cause a very unpleasant experience. I like the idea of filtering them altogether better than providing such bad user experience. It's not ideal - but I think it beats the alternatives. – Benjamin Gruenbaum Jan 2 '14 at 22:47
  • @Shadow Wizard: If there is to be a manual process, I'd rather it apply across all site rather than make moderators on 100+ sites deal with the issue. – Jon Ericson Jan 2 '14 at 22:51
  • 1
    That's not a very convincing argument @Benjamin. Imagine a victim of sexual abuse reading about this on a website where they come to get friendly advice on Judaism. Either the question is appropriate, or not. If it's not appropriate then it's not appropriate anywhere. If it is appropriate then I dislike the idea of automated suppression, just because there's so many things that can go wrong. Especially picking a specific list of words from that particular question. I'd be more up for suppressing this sort of question... – ben is uǝq backwards Jan 2 '14 at 22:51
  • ... i.e. ones that can cause actual hurt, than some that just happen to have a swearword in the title. But then the question becomes who defines "hurt". Someone tried to stop others being called Jesus the other day... – ben is uǝq backwards Jan 2 '14 at 22:52
  • 1
    @JonEricson right, so what about letting SE employees like yourself do that, cross site, with one click? If someone is offended by such ad he'll email the team and the team can handle it quickly. (by the way, for "emergency" cases closing a question should kick it out from hot questions list regardless of votes and answers) – Shadow Wizard is Ear For You Jan 2 '14 at 23:02
  • 5
    @benisuǝqbackwards "If it's not appropriate then it's not appropriate anywhere" I humbly disagree. Some things are appropriate in some contexts in SE but inappropriate in others. Some sites deal with difficult life questions (Like this case in Judaism) and some sites do not (like Stack Overflow). People who don't want to be exposed to the harsher stuff shouldn't have to. – Benjamin Gruenbaum Jan 2 '14 at 23:04
  • 2
    @BenjaminGruenbaum yep, Marco explained it very nicely in his answer below. – Shadow Wizard is Ear For You Jan 2 '14 at 23:05
4

Update:

A reasonable way to do this could be to use "Stack Filter" which is a community extension that adds a profanity filter to StackOverflow and StackExchange chatrooms. It would work quite well for people wanting to avoid trigger words and would not change the site for everybody else.


Old Answer:

While matching the appropriate content to the appropriate site is something that we would all like, it is a very hard problem for a series of reasons.

Here are some examples:

  • Who decides which questions are "representative" of a site and how can it be done? For example, I would argue that questions which are "hot", are hot because they gain traction within their site of origin, thus they are likely well representative.

  • Who decides which questions or topics are "appropriate" on a site and how can it be done?

The problem in itself seems to be no different than this (this is just a hunch): given a question, determine which site it belongs to. It is a very hard problem for humans (see: migration issues, off-topic closures/rants), it is a much harder problem for an algorithm.

Sure, we could filter based on words, but it would be a very bad idea. Word filters are stupid, and for example let's think about the word "rape".

Rape might hurt some sensibilities, but it is not an inappropriate word in any form - surely it represents a violent and despicable act, but it does so in a neutral way.

  • It appears nine times in the Bible, so it can scarcely be out of place in any of our sites.

  • It is certainly appropriate as a scientific term, it appears close to a million times in Google Scholar

  • It is an Italian word which means "turnips" in English, so it's not problematic on that site either.

This word is certainly an appropriate topic on many specific sites (skeptics, philosophy, religion sites, language sites, etc.): how would we determine which subset of the sites it should appear in? Is it appropriate on Stack Overflow?

Let's talk about the word bitch. It is certainly on topic on pets, but in its "female dog" meaning, it's perfectly appropriate everywhere. The fact is: it has two usages, but a word filter won't be able to pick that.

Let's talk about the word dick. It is certainly appropriate as a name. It is certainly appropriate as a cooking recipe title ("spotted dick"). Same goes with the word cock ("cock-a-leekie")

All swearwords are appropriate on English.SE (since it is a language site), but are all questions appropriate? What does that even mean exactly? Do we need to define different sensibilities in an algorithm for different sites?

This is to say: I agree with you in principle, but we need a much better, more fair and adaptable idea to make anything like this work.

  • Very vaguely related (but inspired by this question): skeptics.stackexchange.com/q/18951/96 – Sklivvz Jan 2 '14 at 23:31
  • But this is already being done. Posts that are mostly in a foreign language from the language sites are excluded from the advertisements and hot questions, even though these questions are representative of the sites they were asked on. The current filtering does not account for some topics being "okay" or "not okay" on individual destination sites, which is a crude but simple solution to the questions you raised. – Troyen Jan 3 '14 at 21:35
  • 1
    Relying on people to install a community extension won't help when trying to bring in new visitors to a site (say, Academia), only to have their first impressions tainted by rape questions visible on the sidebar. – Troyen Jan 3 '14 at 21:36

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .