Safety Idea: Reputation Layers

Hello folks, with the ongoing harassment our friends on Twitch are facing, I wanted to spend some time brainstorming what we could do to help protect our streamers. I always prefer to brainstorm with others on the same page, so below is a baseline system we could use to manually & automatically help protect our streamers from hate raids, automated spam, or other unwanted content.

If you are new to Building Glimesh, or Glimesh in general, welcome! This is our Building Glimesh forum and we use it to do all of our work in public. You can read more in our blog post about What makes Glimesh unique.

The Idea: We can add a Glimesh-wide layer of reputation that tracks user activity and engagement on the platform to generate a reputation level. This reputation level can then be understood and used by streamers to make a safer environment for them. The reputation level can also be used by Glimesh systems for automatic anti-spam, anti-harassment, or by 3rd party api consumers for their bots, overlays, or services.

Examples:

A streamer who is getting harassed or otherwise attacked could temporarily require a higher reputation level to chat.

A hate raid coming from mostly Untrusted Users could be automatically blocked, or sent to the mod team for review.

TTS extensions could ignore users under a certain reputation level.

Real world examples:


Levels

All of these levels, their requirements, and their impacts are the start of the idea, and could change over time based on conversations, feedback, and reality!

Level -1 — Untrusted User

Brand new user, no site history, no meaningful conversations. Their account just exists.

Can do:

  • Join streams
  • Chat in streams
  • Follow streams
  • Subscribe
  • Stream themselves
  • Most site functions

Can’t do:

  • Spam messages, some kind of default anti-spam in place
  • PM & other personal features

Promotion:

  • Account has to exist for 72 hours
  • Chat without being banned in 50% of the channels you chat in, for at least two days OR Stream a couple of times without GCT warnings

Holdback:

  • Email unverified

Level 0 — Visitor

A user who has used the platform, but we’re not completely decided on their intent. Most infrequent users will fit into this category.

Jump to Level 0 with:

  • 2FA enabled on your account
  • Linking your Twitter account
  • or, maybe link mobile number?

Can do:

  • Spam messages, regardless of default anti-spam
  • PM & other personal features

Level 1 — Regular

Regulars are common Glimesh users. They use Glimesh with good intentions and they have not violated certain ToS or RoC rules. There is no functional difference between Level 0 and Level 1, except for whatever the streamer configures in their channel. Level 1 being highly trustable, and Level 0 being trust but verify.

Jump to Level 1 with:

ToS or RoC conviction of below offenses invalidate the “jump”.

  • A linked Stripe account

Promotion (within 100 day window):

Unlike the other levels, a user can lose their Level 1 status by not meeting this requirements anytime within a rolling 100 day window. Eg: a Hate Speech conviction would immediately disqualify you for the Level 1 status until 100 days after that conviction.

  • Must chat in 5 unique channels OR participated in at least 10 hours of streams

  • Must not have been convicted of any of the following ToS or RoC violations:

    • Harassment
    • Hate Speech
    • Threats & Harm
    • Spam or Unsolicited Advertising

    Note: ToS & RoC enforcement is independent of this system, and users who have violated the rules will have their account punished per our GCT enforcement table.


Automatic Moderation Systems

We could build out automatic methods of detecting high system usage from untrusted accounts, especially to channels they don’t normally interact with and automatically block them.

We could also implement basic rate limiting based on your reputation level. This would prevent system-wide ads or spam to all channels.

Channel Options

At Glimesh we believe that the streamer should have the ultimate control over their channel, so we don’t want to enforce too many restrictions by default. Instead we give the streamer knobs and buttons to customize their channel to their liking.

Required Level to Chat

A streamer configurable minimum level for all users in their chat to be able to send messages.

Required Level to Use Links

Require a certain level for the chatter to post links.

Required Level to Use Emotes

Same thing but for channel / global emotes.

Required Level to Use Clips

Once we have clips, require a minimum reputation level to use them.

Required Level for Follower Alert?

I don’t think this would be something we implement, it would likely be something for our API users to use.


API Integration for App Creators

This would allow app creators to setup the same granular controls for their notifications, TTS, and other features that are easily abused.


Thank you for reading the idea. I'm looking for any and all feedback from anyone who's willing to give it either in this Building Glimesh forum, a Discord call, or whatever medium is comfortable for you.
4 Likes

Great idea. Question/suggestion streamers Mods. We currently are able to set our mods. Example: new streamer joins glimesh. Streamer brings their fan base to glimesh including their selected Mods. Streamer sets mod roles but. Base off reputation mods would have limited access due to lack of reputation. I.e. sending links for streamer, clips (when we get them), etc.

If reputation would be a thing, I as a streamer have trusted mods/viewers that help my channel. Limiting their access I feel would make them think 1. I don’t trust them 2. The platform doesn’t trust them.

Reputation features should work as a overall platform mechanic but also individual streamer channel mechanic where the streamer based on roles could override platform rep options via role management.

But I still think it’s a great idea.

4 Likes

Good point! I think mods would be exempt from this inside the channels they mod for, same thing as streamers, GCT members, or admins.

1 Like

Playing devils advocate also. How would we market it to the public and other streamers. I think some viewers my be turned off to the fact they would have “earn” the right to use features that they have on other platforms.

Streamers with “open q/a style chats” where a user might send a link to source material might not see the rep system as a good thing.

Sorry bouncing ideas in my head.

1 Like

Ultimately for a streamer, it would work exactly like these options do today:

Which are marketed as configurable features streamers can use to protect their stream. These levels would enable the same thing, but with a bit more magic. Hopefully it’ll enable streamers to have more open chats, without having to deal with targeted harassment.

For users who are impacted (for example Level -1 users), they simply need to link 2fa, or maybe their twitter account, or some other verification to continue chatting. Each of those features also helps protect the security of their account.

If a streamer wants everyone to post links, regardless of their level, they would be able to do that.

1 Like

For hate raids, you state sending them to mod team for review. Could it be done later?

After the raid is in a mod could “unraid” it. So anyone who came in from the raid is muted/booted?

1 Like

Do we think there is a viability to being able to quickly turn some of these options on with just 1 command. Sort of like a panic button that someone can hit/type in the event of a hate raid type situation which turns on a bunch of these features that maybe aren’t always needed for the creator.

For instance it could turn on the ability to chat or post links from level -1 to level 1.

Sounds possible to me, but we’d have to understand the impact if it’s not handled in real time.
For example, a real positive raid could be blocked automatically and not seen until sometime later.

I think that’s exactly the kind of thing we could let them control. “Hey there’s an influx of raiders with zero reputation, should we block the entire group?”

Sounds like a good idea to me! Proactive & reactive solutions!

Not to be an ass here, but I would assume this is supposed to say “can’t do”

From what I have seen as well, people seem to be getting round a lot of word filters by using things such as special characters, it could also be worth (if its not something already possible) looking into how to stop the use of special characters.

Streamers to be able to implement their own banned words may also be a consideration here.

I agree with the idea of a quick command as @Chalk has suggested to quickly toggle these things on, I would also look if we could extend that too mods, as if a streamer is invested in a game they may not notice it to react immediately.

Perhaps a way to turn off chat entirely, preventing users chatting at all, regardless of level (mods to override this ofc)? Like a killswitch almost, this is happening - stop all interactions.

For real time I think for the numbers these raids bring some form of mass click and ban for the mod team may prove beneficial.

Just initial thoughts here, not sure if it is of help, as I think more on things I will come back.

I think getting a start on this now and pushing forward can only be good for the future of the users of this platform.

Can is correct, Level 0 can do all functions. Level 0 and Level 1 have no functional differentiation, just allows the streamer to basically select “Random Viewers” vs “Glimesh Regulars”.

What I’m suggesting with Reputation Layers goes beyond the contents of the message, and instead focuses on the reputation of the user. Users who join the site only to hate raid, spam, or otherwise automatically harass users would be stuck in Level -1 (along with being banned by the GCT).

While I agree filters like these are a good idea, they are not directly related to the proposal here.

One of the main things we’d hope to accomplish with the Reputation system is to allow honest users to continue chatting. I’ve been watching what happens on Twitch, and a streamer losing their chat to prevent spam is not a solution we should shoot for. Ideally we can design this system in such a way that a streamer can continue about their business, not even aware a hate raid is happening, because we’re automatically blocking it.

Oh I see, the use of 2 level 1’s, (i missed the minus) one for Untrusted User and one for Regular threw me, I apologise for the misunderstanding :slight_smile:

Love all this! A little like Mixers old Viewer levels per stream but this is site wide, which makes more sense. Yes please :relaxed::joy:

Maybe that’s the nerd in me! We can change the numbers, or words attached to the levels :stuck_out_tongue:

Would there be a way to lose reputation. Say someone comes in, they enable 2FA or whatever, and they cause issues, could their reputation level be reverted based off any infractions made etc, so they would have to regain that trust again via the interaction levels sat at -1 ?

My logic in that is that not all people who are hate raiding are people who are potato accounts, they could be users who have been round, even streamed etc in the past.

I like everything but this bit:

Required Level to Chat
A streamer configurable minimum level for all users in their chat to be able to send messages.

If I was a new user, and I joined multiple channels where I was not able to chat, I would find this very frustrating, and I would probably leave the site. I can see a lot of streamers setting this to “glimesh regulars only” without thinking about it (eg. in an effort to avoid bots). We want visitors to feel welcomed, not shunned. I understand it is a choice the streamer is free to make, but if it hurts the glimesh community as a whole by putting people off about the platform, I’m not sure that’s a good idea.

If we did implement it, I would recommend we limit the streamer’s options to

  • -1 untrusted and
  • 0 trusted visitor

on the other hand… is it really that different from “sub-only” chat or “follower-only” or “emote-only” chat? :thinking:

I think the idea is to be less restrictive than " follower only" chat but also potentially more effective at sussing out bots and troublemaking users.

would “spam messages” be defined as a message identical to any of the last 10? ie. not just the same user reposting the same message, or sending messages rapidly, but any non-unique message from recent chat

(feel free to spin this off into its own thread if you think it’s warranted)

Potentially could be both. What I would envisage the aim of stopping/allowing this spam would be things like raid messages (both the good kind as well as bad) where the same message is generally posted by multiple users multiple times.
So essentially someone that isn’t allowed to spam because of their trust level would be stopped because of either

  • too many messages too quickly (slow mode style)
  • non-unique message (same message as everyone else above)

The second one is fraught with danger as it could easily pick up situations that are not the target, such as a streamer asking a question (What state is Las Vegas in again?) and everyone in chat replies with “Nevada” within a short period of time. This isn’t really something that is an issue of concern, nor do we actually want to pick up this with any sort of spam filter.

Restricting it to just that users last few messages is probably a better option. If its only people with a ‘poor’ reputation that can’t repeat the same message then I believe its more likely to have the desired outcome.

yea I was speaking of spam filtering in the context of reputation levels - spamming would not result in a ban or timeout - the message just wouldn’t send if your reputation level was not high enough

Had a good read through of this idea.

I really like the idea, I feel a sort of Panic Button is a good idea that will temporarily up the reputation level to chat is a great idea. I also dont feel it will be something intrusive like how people used to lock their mixer chats down unless you had reached viewer level 10.

I see this system as one that triggers when spam is detected or when the panic button is hit. Not something to stop new users talking. I really feel if this is done well it could really be a good selling point for protection against hate raids.

1 Like