Assessing, clarifying and improving member bans and ban appeals

[Arlodotexe]

Background

From https://discord.com/channels/725513575971684472/929868175762395146/1328414399844716637

A brief assessment of our infraction system

Posted on Discord in #rules-and-info:

---

🛡️ Infraction policy

• Breaking the rules may result in an infraction.
• When a strike or warning expires, all infractions are removed.
• Infractions stack. If you had a Warning, you now have Strike 1, and so on.
• Bypassing the infraction system or violating a rule shortly after joining will result in a ban.
• Certain offenses can be pardoned or made more severe, at an Admin's discretion.

---

• Warning - Lasts 2 weeks
• Strike 1: 7 day mute, 21 day strike
• Strike 2: 21 day mute, 63 day strike
• Strike 3: 63 day mute, 189 day strike
• Strike 4: 189 day mute, 567 day strike

---

Server Companion#4732 is used to enforce our infraction policy, open source here:
https://github.com/WindowsAppCommunity/uwpcommunity-backend/blob/dev/src/bot/commands/infraction.ts#L267

---

How 'bans' currently work

We don't ban, except to combat bots and geniune malice, which usually manifest under specific conditions:

1. A member breaks the rules immediately upon joining.
2. A member purposely circumnavigates the infraction system.

Clarification request

Can members be unbanned?

In short, we can review case-by-base, but we have no appeals system in place because we don't usually ban to begin with, especially not established members.

Our infraction system is designed to give second chances to honest (non-malicious) mistakes so people can learn and grow from them, but it doesn't tolerate reckless disregard of rules and community.

Even if malice is intentional and short term, circumventing the infraction system and breaking rules on first join are the only two things we ban for. For established members, we don't ban.

Problem

This is less than ideal for an established member, even one who's acting out.

Again, our infraction system is designed to give second chances to honest mistakes so people can learn and grow from them.

We don't tolerate genuine malice or blatant disregard towards the community or its members, but we must also recognize and encourage the potential in each individual to learn from their mistakes.

Accounts that exhibit botlike behavior are a different story, and shouldn't expect any changes.

Solution

To improve our process and provide this clarification for the broader community, there are several ways we can work to improve this. Each will need proper analysis and mapping before implementation anywhere.

"Just reapply the roles"

It's worth noting that even trying to leave the server to reset your roles is an indicator of malice. This indicator is a useful signal to moderators that this person is actively trying to be a problem, which is why "just reapply the roles using the bot" isn't useful unless mods can be notified about it.

Unfortunately, our bot is currently stuck running the code it has.

We're actively working on our new bot, which is where "just reapply the roles" can be implemented in a task tangential to Implement infraction command WindowsAppCommunity/WindowsAppCommunity.Discord.ServerCompanion#6. This ticket needs some minor planning before opening.

Clarify and improve conduct process

Ideally, we shouldn't need to build an appeals system at all, because real humans (especially established members) should never get indefinite bans.

• Malicious accounts run by bots get the ban hammer.
• Malicious accounts run by humans get infractions (warnings and strikes) where the mute/strike duration for each tier is 3x as long as the last.

Rather than an indefinite ban, when a human tries to dodge the infraction system their ban should only run for the duration of the mute. If the member returns, the strike should be re-applied automatically.

Unfortunately we're not able to automate this process, so we'd need to case-by-case review each person that wishes to return until we've built that. This is more or less the idea of an appeals system, but we don't call it that here due to the informality.