This post has been de-listed
It is no longer included in search results and normal feeds (front page, hot posts, subreddit posts, etc). It remains visible only via the author's post history.
edit: This went live for all communities on May 5th, 2023
Last August, the Safety team posted an update on the Ban evasion filter, a mod tool that automatically filters posts and comments from suspected community ban evaders into the modqueue. We are happy to announce that the tool is being released to all subreddits over the course of the next few weeks! Once live, we will let you know directly.
How does the feature work?
Ban evasion filter is an optional subreddit setting that leverages our ability to identify posts and comments authored by potential ban evaders. We identify potential ban evaders based on various user signals related to how they connect to Reddit and information they share with us. Our goal in offering this feature is to help reduce time spent detecting ban evaders and preventing the negative community impact they have.
Once this setting is available to your community, you can find it by going to Mod Tools -> Safety (under Moderation section) > Ban evasion filter. When the setting is turned on, you can set your preferences on how much content is filtered to the modqueue. The preferences include:
- Time frame: which allows you to set a timeframe for how recently a user was first banned from your community. FWIW, our data shows that communities tend to receive content more negatively from users who were banned more recently.
- Confidence: which allows you to set a leniency threshold for posts/comments separately.
Settings for the Ban Evasion Filter
When content is filtered for ban evasion it will show up as follows in the modqueue:
A comment filtered by the Ban Evasion Filter in the modqueue
Note that when we roll out the feature, it will be âoffâ for all communities, and you can turn it on at your discretion. The exception being communities in our Beta, who should not see any changes to their settings.
Limitations
While we are really excited to make this tool publicly available, there are a couple limitations to be aware of:
- Accuracy: It isnât 100% accurate, as the user signals we use are approximations. Please use your discretion when deciding to allow users to participate in your community. If a positive contributor is getting repeatedly flagged, know that you can prevent their content from being filtered by (A) adding them to the âApproved Usersâ list in your settings, or (B) manually approving their filtered content three times.
- Latency: If you unban a user and in the following few hours they begin engaging again by posting or making comments, the ban evasion protection filter may still flag posts or comments from the recently unbanned user and place them in the modqueue. Once the system updates to identify that you approved them, they should be able to engage with no issues. This is just one example of latency that has prevented perfect performance, but as you use the tool you may notice other examples.
Also, please note that if you were a participant in the Beta communities, our most recent updates will not be applied retroactively to content that was previously filtered by the Ban evasion filter. As we continue supporting the portfolio of safety tools for moderators, we will work on making this one faster and more accurate without compromising on privacy.
Whatâs next?
We know there is more for us to do. If you suspect ban evasion in your community that we may have missed, please file a ban evasion report using the /report flow. Note that your reports and your usage of the filter informs how we detect and action bad actors. We will also be continuing to improve the signals that inform ban evasion detection.
Before we goâŚ
We wanted to thank our Beta members. Our Beta communities have been amazing at delivering helpful feedback that inspired feature improvements such as details around recency and adding more clarity and granularity in the settings page. Thank you once again to all the communities that participated and passed along feedback.
We know that this has been a challenging issue in the past, and so we are excited to make some headway by making this tool available to all qualifying communities. If you have any questions or comments â weâll be around for a little while.
I am very curious to know what privacy concerns would apply to mods using the new signals via 3rd-party tools, but not to mods using the official apps/website where the same data is visible?
This just feels like yet more restrictions against mods who prefer 3rd-party apps and tools over Reddit's own.
As a mod I must say this is a lifesaver against spammers and vrious kinds of regular offenders. I beta tested it and deployed it on most places I own under various accounts. I definitely believe that you shouldn't post where you're not welcomed, and coming back after a ban is the best way to harm yourself. If people don't align with you, just let them be is my motto...
But I am also a Reddit user with separated accounts operating in various communities, and more than onces I've seen fringe to completely abusive moderation practices like banning users that are active in certain competing subs, mods banning unverified commenters when their rules clearly states you should be approved to post a new thread... I've also reclaimed several abandonned subs and know several others who made it, I can easily imagine someone doing it and banning several users on all of those... I've seen various collusions that were designed to harass someone, mass reporting any post they make...
So I'm a bit worried about further improvements that may have the potential to be misused.
Subreddit
Post Details
- Posted
- 1 year ago
- Reddit URL
- View post on reddit.com
- External URL
- reddit.com/r/modnews/com...
Will there still be a means to provide feedback after the official launch, and if so, how?
As I'd explained in my past feedback, we've long had three of our own ban-evasion detection tools that we had hoped to benchmark against Reddit's built-in one; but the only community of mine that was ever selected for the beta was one that gets no posts (and therefore no ban evasion), thus we never got that opportunity during the beta period. Will we still be able to do that even after the wider roll-out?
Additionally, the continuous updates we've been getting have mentioned a new confidence level, but talking with others who have the feature on communities where they could actually make any use of it, it doesn't appear to be exposed in any way to the API. Is this something we'll be able to read from the modlog entries via the API, or can that be added? That seems like an important data point to us, so we'd like our tools to be able to make use of it.