Create Automod rule sets based on viewer loyalty
Make rule sets of automod strictness, whereas each set is applied to a type of viewer determined by streamer. For example:
- Followed less than a week ago: Very strict.
- Followed more than 6 months ago: Less strict.
- Subscribed for 5 years straight and <2 timeouts: No auto-moderation.
The main problem with a universal automod rule set (i.e. current automod, as of Dec 2021) is that when a flaggable word is used with amicable intent by a loyal viewer and then manually permitted by a mod, the same permitted word can now be used maliciously by bots and unsavory viewers.
This simple solution introduces a gradient more representative of realistic viewer behavior, thus reducing false positives. I'm certain Twitch has both the data and the technical staff to make this invaluable differentiation possible in a short amount of time.
After all, it makes no sense that both of the following messages are treated the same way by automod:
3-year loyal viewer: "Are you %@#&ing serious?? Congrats on the new car!!"
New account, followed 5 seconds ago: "%@#& your ugly teeth"