I got forwarded me a message from a user who was (IMHO understandably) taken aback by his relevant and useful post being flagged as spam and especially the accompanying message they got by e-mail:
Your post was flagged as spam: the community feels it is an advertisement, something that is overly promotional in nature instead of being useful or relevant to the topic as expected.
I agree that the wording “the community feels” is quite inappropriate for a bot flagging something as spam. It seems that the message has to be the same for both automatic flags and user-made flags. I have found and edited the message for notifications about spam flags to:
Your post was flagged as spam by either a human user or an automated system. This usually means the post has some features of an advertisement or undesirable promotion.
Does that look better?
Additionally the reason for flagging as spam was:
This new user tried to create multiple posts with links to the same domain.
All posts from this user that include links should be reviewed.
In my experience this does happen somewhat frequently (when I was more active I would certainly see an innocuous post flagged for this reason about once a month) - people create a bunch of posts with links to manuals etc.
Currently, a new user (Trust level 0) is flagged as spam if they create 3 topics with links to the same domain. I think that’s a sensible value, but I’ve added some of the frequently mentioned sites to whitelist. (notably: mc-stan.org, github.com, github.io which seem to cause a majority of thous problems) This should alleviate the problem.
Note that users quite quickly loose Trust level 0 - you just need to read a bunch of posts and spend some time on the site, so this shouldn’t affect too many users anyway.
Any feedback on the changes welcome.
The new wording looks better to me!
I suspect the flagging (was/is) based on messages of any types: replies or topics. I seemed to run afoul of this system when I replied to my topic with a link I had already used earlier. I was told that I could not use a link to “that domain” (though oddly it did not tell me what the offending domain was, so I just removed all links) but I successfully did so earlier in the topic. Two of my posts in the topic were flagged shortly after that, both of which had that link. If this is too specific and you’d like to obfuscate the inner workings of the system, feel free to redact this.
It might be helpful to have a stickied “rules” post that includes a line about new users going easy on using links in their first posts. If that already exists then I had missed it. Even if they don’t read it, at least it can be linked to in the “your post was flagged” message to let them know it wasn’t arbitrary and they could have avoided it if they read the rules first.
Proposed wording for such a stickied notice:
“This community uses an automated system to combat spam. If you are a new user, please be careful to keep links to a minimum to avoid having your posts automatically hidden. You will be notified if this happens and it can be unhidden by a moderator pending their review.”
You are right, the limit is on posts not topics. I think the system misfiring is sufficiently rare to make an explicit notice to new users not worth the added noise (based on anecdotal evidence, most new users do not read most of the guidance the site provides anyway). Most commonly this was triggered by links to Stan manual or GitHub issues, so with those domains now on the whitelist it should become even rarer.
Note that to trigger the system, you need to be at trust level 0 AND post repeatedly with links to the same domain. Trust level 0 is lost by reading at least 30 posts in at least 5 topics and spending 10 minutes on the site (you became trust level 1 - basic user shortly after posting your inquiry). At the same time, the site should already encourage you to edit a post instead of chaining multiple replies by the same user which would usually avoid the issue. I don’t blame anyone for not reading/following that guidance, but it IMHO makes it unlikely that more guidance will help :-)
And we also do get some real spam, so I am not to keen on softening the rules substantially.
Does that make sense?
Thanks again for bringing the issue up!