Just gathering ideas.
Automods are bad tools that result in more abuse than the problems they solve.
I made a post on this some weeks back. Whether it would be through an ‘automod’ bot (ideally not, in my opinion) - or community settings preventing posts from even going out is another matter.
- Automatic removal option to remove posts and/or comments for specific keywords. This would be most useful for automatically removing posts and comments when people slur. Piefed already has a keyword filter for visibility. This could be expanded to community settings. Have it also fire-off a report to the moderators when someone triggers it.
- Automatic URL removal. Allow communities to blacklist specific urls. Useful for politics or news communities that want to negate sources known for misinformation.
- Automatic removal for repeat URL posting. Very useful for politics or news communities to prevent double-posting.
- Make it so a community can set itself up to only accept text posts, video posts, or image posts. This should prevent tedious janitorial cleanup for communities that only allow links, or text posts (the most common two).
- Post Delay Restrictions. Some communities, perhaps not many, might be interested in posting cooldowns for users. So you can only post 1 post every hour, or 2 posts every hour - or whatever the chosen limit is. This would help negate spammers and over-enthusiastic posters flooding a topical community.
- Post Formatting Requirements. This one could be trickier and more effort than most of the others, but setting conditions for the formatting of new posts would be useful.
All of them are possible with my current design plan, which is allowing mods to write the functionality themselves (instead of predefining them like in some other automods) in sandboxed Lua (might change).
I might still predefine some myself for performance.
Scheduled posts
In the mean-time this works https://schedule.lemmings.world/auth/login
Piefed has them
For automated ones, sure. But if you just want to schedule some post you made, that functionality is already available in Piefed and will be available in Lemmy in the next 1.0.0 release as well.
Piefed has scheduled posts. I haven’t used it because I have no reason to, but the functionality is there.
Automatic comment replies based on keywords or key phrases.
Basically, if a comment contains a specific string, the automod would reply with a corresponding string, configurable by the community moderators.
Would this be doable?
Duplication post removal.
Sometimes two or more users post independently on a short time not knowing that a thread already exist.
For context, this is what reddit’s limited automod is like
https://www.reddit.com/r/reddit.com/wiki/automoderator/full-documentation/
I’m sure we can do better. For example, being able to use variables
Of course. Reddit’s automod is very simple.
My current design plan is leaving the implementation to community mods by embedding sandboxed Lua (might change).
Set a minimum account age for posts and comments.
- Controllable by a simple, easily understood scripting language. Lua for example.
- A sense of time and date.
- Powerful regex interpreter.
- Open source.
- Documentation. Maybe a wiki?
- Easily configurable for each instance so that admins aren’t afraid to install it.
Automod was one of the worst things to happen to Reddit from a user experience standpoint. So what I want from an automod is to never see one.
Without automod functionality, every single community of notable size becomes utterly bombarded with off-topic posts and spam.
Unfortunately there is also a need for automatic moderation.
Humans can’t catch everything.
By the way, mind telling me what was the most infuriating thing about Reddit’s automod to you?
Automatically removing comments without human oversight and for mundane things.
Randomly triggered a hidden keyword? Goodbye post.
Only wrote 49 words instead of 50? Goodbye post.
Account didn’t get enough updoots from strangers prior to posting? Goodbye post.
Call me old fashioned, but I would rather a human at the helm of such decisions not an algorithm.
This would just burnout moderators on a highly active communities. My preference is not have these tools work after-the-fact, after-the-post but simply tell the would-be poster that their post has hit a keyword block. But basic stuff like mandating all posts be link posts or text posts (depending on the communities focus) or instantly removing duplicate posts seem pretty necessary for many communities if the fediverse expands in population.
Account didn’t get enough updoots from strangers prior to posting? Goodbye post.
Now this one, I admit is a tough one - as it can be harsh to new users. But it’s simply based on trying to deal with spam posters. As the Fediverse grows, the high-trust public nature of downvoting, public post histories and public mod-logs should negate people wanting this.
It’s a risk I’d rather take, than letting automation make this another reddit where all the big communities are managed by algorithms.
A human could see a post was only 49/50 words and apply their judgement to know that this post is acceptable because the quality of what was said was more important than the quantity.
A human could see the word trumpcard was in fact not about Trump.
If the price to pay to be moderated by humans and not algorithms is that obscenely large communities can’t exist, than we should be pushing for human sized communities.
It’s a risk I’d rather take, than letting automation make this another reddit where all the big communities are managed by algorithms.
That’s not fair on other people. It’s also just not viable. I doubt almost any moderators would support not doing this.
Moreover, there already exist self-hosted bots that already autoremove content on the Fediverse. It is already happening de facto.
A human could see a post was only 49/50 words and apply their judgement to know that this post is acceptable because the quality of what was said was more important than the quantity.
I wasn’t specifically talking about that kind of automodding. I’m not sure if that is needed, except maybe in specific long-form debate/discussion communities. I don’t think that level of automodding is a priority, to be sure.
A human could see the word trumpcard was in fact not about Trump.
Yes, there could be false positives. I’m sure that if a community did have a “Trump” filter they could specify it only to notice “Trump” when posted as a single word.
But I was also more thinking about autoremoving slur posts and comments here.
If the price to pay to be moderated by humans and not algorithms is that obscenely large communities can’t exist, than we should be pushing for human sized communities.
A community would not even need to be obscenely large in order to not become a moderating chore without some level of automod functionality.
Welp, the fourth one can’t happen. Posts or comments by users won’t be stored in the database, so you can’t calculate “karma” for users.
The others are left to the community mods’ implementations.
I would like to be notified when users who are on a warning post or comment.
