Auto-immune heavy-handedness: Anti-spam rules can appear to be overzealous community flagging, stymieing community

Hi all. I run several Discourse instances, so have a reasonable amount of experience with Discourse from a moderator’s/admin’s perspective. Recently I joined a new community (as a brand new user) running a Discourse forum and made some posts in an active topic in which I included some (slightly) confrontational points and, in response to other posters, a post with a number of links further explaining my position. My 3 initial posts were “flagged by the Community” (or similar language), and marked “ignored”. I received a notification pointing me to the “community guidelines”… (I couldn’t see how my posts infringed them). My post with the links was disallowed by Discourse, claiming it contained too many links. I tried a new tack and created a blog post off the Discourse instance to contain the references and posted a link to the blog post in a second attempt at a response. I was informed by Discourse that my blog was “a blocked site” and my post couldn’t be saved/posted.

Despite my Discourse experience, I got the impression from the specific wording of the Discourse warnings, that a) my posts had been flagged (i.e. seen as offensive/against the community guidelines) by human community members and that b) my blog URL had been blocked by a human moderator. This left me feeling quite angry and unfairly treated by the whole site and its community.

Turns out, it was a couple of days before any human moderators reviewed the situation, by which point, I’d developed a serious sense of hostility towards the site… they informed me that the flagging, the pointer to the community guidelines, and the block on my blog URL were all automatically carried out by Discourse algorithms. No humans involved. They determined that I had not, in their view, infringed their community guidelines, and they reinstated my “ignored posts” and unblocked my ability to link to my blog. They also increased my trust levels to reduce future problems. They also apologised for my negative experience (and agreed that they, too, would have found it unpleasant if they’d been in my position).

I found the wording of the warnings from Discourse ambiguous. I’d be keen to see the anti-spam features and the reporting carried out by Discourse algorithms tweaked to make it very clear that such measures are triggered by automated anti-spam algorithms, and are non-human interactions, to avoid such ill-feeling in the future…


It looks like you were linking to the same external domain (URL) many times?

That would definitely trigger what you’re describing, when a new user creates a lot of new posts containing the same URL. Especially if this happens in the first 24 hours after account creation.

You’d need to provide a little more detail for me to be sure.

Yes, that’s the case - I was pointing someone wanting to use FOSS tools for their community at specific posts on my blog… I realised later that the post was blocked due to the use of so many links in a post by a recently joined user… that part isn’t the problem - it’s the ambiguity of the system message, which I initially understood (along with the system messages telling me about my “ignored” posts) that made it sound like my site and posts were blocked due to complaints (flagging) by a community member(s)…

I found the language of the messages ambiguous regarding the actor - I only later found it was completely automated algorithmic behaviour by the Discourse system and not site users or the site’s moderator…


I understand, and thanks for bringing it up. It’s a balancing act between protecting the community from spammers, and allowing new users to post what they want to post.


I’d agree this is a good idea, and there should ideally be a way for users to easily flag any action they were subject to for review.

1 Like


I think Discourse does a great job of “striking the balance” between detecting and flagging new users posts. Spam is much less of an issue (and takes far less time to deal with, and less custom code to maintain) than on our legacy vB forum.

Yes, we often see new user’s post flagged because the copy-and-paste some code or log file info into a new post, looking for help, and the post if flagged “typed to fast” or something similar, but we consider this a feature, not a bug.

Detecting and managing spam is a non-trivial task, and I think Discourse has done a good job, as thing stand now.

I think the wording is very important. Big companies also mess this up. I recently wanted to use uber eats, but because I stay abroad, their systems flagged me. But it was more like a shadow ban: there was no mention of spam protection mechanisms, just generic error messages. So I started spamming insults into their “support section”. I thought no one would ever read this, but I actually got a response and finally my food. They also clearly had a process for people to get unblocked, but it wasnt clear how to reach it from the UI. I think this dystopian newspeak about “community guidelines” should generally be avoided. It should just say: you were caught by a spam protection mechanism, here is a captcha to prove you are not a spammer/ please write a personal message that is send to a mod explaining why this is not spam.
this “community guidelines” stuff coming from big tech without clear recourse and statement which rule was broken is very nasty. I think this will lead to their eventual demise and its best to stay as far away from this as possible. Conceptually and by language.


I can agree that on some cases, the message to the user should be more descriptive about what actually happened.

Transparency in moderation is really important so that users know if they did something wrong and what exactly was that. So, I think there should be a distinction between the messages sent when a post is automatically flagged vs when users flag it. Also, a next step should be provided to the user to appeal automated decisions (eg: pm a moderator).

In the case of TL0 flagging, this is especially needed because it impacts new users that are not familiar yet with the discourse system and might get frustrated with such ambiguous messages.