Feature Request: Content Warning Categories

[dtschust]

Content warnings are a very useful feature, but they don't currently provide much customizability for the user reading toots with content warnings. Apps usually allow you to either hide all content behind content warnings, or show all content behind the warnings by default. There's no granularity here.

Personally, I would like to automatically show all content with warnings about programming, but I do want all content with political content warnings to remain hidden.

In theory, an app could provide a user with a regular expression or string to automatically show all "programming" content warning posts, but since content warnings are free text entry by the poster, this feature would not work very well. Posters could spell programming wrong, write "coding" instead, or "javascript", etc.

The feature I'm suggesting is:

• Allow an (optional) content warning category section for all new toots. This exists alongside the free text entry "Write your warning here", so a user could select a category like "programming" but still add their own specific warning, like "CSS Drama".
• Vanilla mastodon would have a fixed list of preset CW categories. Here's an incomplete potential list of the top of my head: [ "Politics", "Spoilers", "Trauma/Abuse", "Joke Punchline", "NSFW", "Programming", "Mastodon Meta", "From BirdSite", "Other" ]. Other would be the default, and would be ignored and not actually sent. It's the same as not including a category.
• Individual mastodon instances could add their own custom CW categories, and perhaps disable some of the vanilla ones. For example, a sex workers instance could remove the "NSFW" CW category because they allow NSFW content freely on their platform. Similarly, an instance that, for example, requires an "eye contact" CW warning for any picture involving eye contact can add that category.
• Posts with a category set for their content warning would have that category displayed, as well as the "Write your warning here" value. If you have a content warning of Politics with the custom text, GOP, the content warning would look like "Politics: GOP". If custom text wasn't defined, the colon could be dropped, i.e. "Politics"
• In settings, users can then toggle auto-hide or auto-show for each individual CW category, and the client will honor this setting when displaying toots.

---

• I searched or browsed the repo’s other issues to ensure this is not a duplicate.

[holyspiritomb]

That would be a really cool feature if implemented!

[rhoio]

how would you propose this feature to work with localization?

[dtschust]

@rhoio that's a really good point! I'm not familiar enough with the product to know if we provide any other localization efforts for instance-specific fields.

For example, do we allow translation of custom emoji descriptions?

For the core supported categories, we could definitely provide translations, and perhaps separate out a displayName that is translated and a value that is always in English to make it easier for apps to display a translated value but still filter out all "Politics" categories, regardless of language.

FWIW, I believe content warnings themselves have this issue, as they are user generated already so there are no translations, and they appear in the language they are written in regardless of the preferred language of the user viewing the toot.

[rhoio]

Content warnings as they are now don't have the issue because the author of the toot chooses the language according to their audience (if they write in japanese to a japanese followership, use japanese cws etc).

Making CWs part of the interface creates the translation problem. Maybe instead of allowing instance admins to add custom CWs, simply provide a wide set of CWs and let instance admins select from this set which CWs they want to offer.

I feel a bit ambivalent about this proposed feature. It would be yet one more item in the status form which is already a bit overwhelming for new users with all the offered possibilities. On the other hand the standardization would improve filtering possibilities...

[rugk]

This is really a great idea, because it solves one great problem I have with Mastodon: People use CWs for everything!!
As such, I e.g. just routinely click on CWs, because in 90% of the cases it is totally benign (I e.g. do not care about programming threads or so). However, I may really not want to see NSFW content or so.
And as people do not always label CWs at all, or if they do, the descriptions are not very helpful, I cannot see from the outside if I am about to unhide a toot I want to see or not.

That's just my use case.

[tastytea]

How about we stop using the summary field for CWs and use tags with a special type for content warnings? The server could have a database with all the previously used CWs and how often they've been used, possibly with an optional synchronization mechanism between instances. This information could be used for auto-completion.

Example: If you type “cat”, a little list pops up showing that “cat” was used 17 times and “cats” was used 2 times. You see that most people expect cat pictures to be labeled with “cat”, not “cats”.

This would also work around the problems with localization and instance admins wouldn't have to maintain a CW list in multiple languages.

Users could set a default setting for CWs and maintain a list of exceptions. For example: Have all posts collapsed by default, don't show posts labelled “Twitter” at all and show posts labelled “Programming” expanded. It would be easy to add boolean logic in an advanced mode.

I wrote about it in more detail in this blog post.

[ThisIsMissEm]

Given existing UIs, I think we'd first have to start publishing an aggregate of common content warnings (perhaps like hashtags instead of arbitrary strings), and then clients could query that API endpoint to suggest warnings to use

i.e., you need some way to get from free for all warnings to communities settling on warnings they use

Then you'd have an easier time to implement features like "ignore warnings for X"