JAIHUB Wiki Archives

You've loaded an old revision of the document! If you save it, you will create a new version with this data.

===== Proposed Moderation Changes =====

The following are proposed moderation changes. Whether these will be implemented, changed or updated remains to be seen.

===== Introduction =====
This is a curated list of suggestions provided from the community of former and current moderators for Janitor AI. The objective of this proposal to the development team is to collaborate on improving the experience of moderation, collaboration, and developing best practices moving forward.
===== Site Moderation =====
==== Bot Visibility ====
Develop an additional visibility setting for profiles and bots to allow moderation to view content without it being visible to users. Currently, bots can be either public or private. Once private, a bot is not visible by moderation. 
==== Code Changes, Genuine Mistakes ====
An internal view setting allows for moderation to remove a bot from public view without restricting moderator view. This is ideal for allowing creators time to fix issues with their bot while preventing TOS-violating content from appearing to users.

Removes content violations from public view while still being visible for moderation purposes. Potentially trending time could be paused and resumed once the content has been adjusted sufficiently.

==== Assistance with Private Bots ====
When the automated moderation tools prevent a user from publishing a bot, moderation is currently unable to assist effectively due to visibility restrictions.

Users could opt to toggle their bot as ‘visible by moderation’ or generate a moderation-only link, then attach to a ticket for support.

==== Manual Approval ====
When automated moderation tools prevent a user from publishing a bot, a moderation-only visibility would allow moderators to manually approve or deny a bot publication.

==== Media Restrictions ====
Implement restriction options on inserting image media into bot descriptions, profiles, or bots.

There are some chronic offenders with more than 50 infractions of violating the image guidelines. To prevent inappropriate media from being visible by users, it may be useful to restrict chronic offenders temporarily.

==== Restrictions on Insertable Media ====
Loss of the option to embed images in bot descriptions.

==== Restrictions on Profile CSS and Media ====
Loss of the option to modify CSS on a user’s profile, in the case of profiles that abuse CSS to hide user tools (block/follow/report buttons).

==== Restrictions on Bot Images ====

Loss of custom bot profile images. Instead, a placeholder is automatically picked. Alternatively, a guideline-approved gallery of images could be used to select an image. This would allow restricted users to still have appealing bot profile pictures without the possibility of potential violation.

This is only for chronic offenders. A threshold and circumstance set would need to be set by site development and staff.

==== Chronic Offender Processing ====
A streamlined flow of structured disciplinary actions to follow grants a clear cut path to permanent bans.

Example: A user creates a guideline-violating bot for the first time. This is their only infraction. As a result, they receive a warning and their bot is deleted.

Example: The same creator ignores the warning. They make a second guideline-violating bot. The bot is then deleted, a warning is given, and they are banned for 3 days.

This proceeds to a 7 day ban, then 14 day ban, until a permanent ban is reached. This process is applied to every user fairly, regardless of follower count.

Immediate bannable offenses are made clear. For example, CSAM is an immediate ban with no warning, as it is egregious and neglecting TOS is no excuse.

==== User-Flagged Content ====
Content that receives a specific threshold of user reports within a set timeframe (e.g. 10 reports in 12 hours) should be flagged as needing urgent moderation. This prevents guideline-violating content from going unnoticed in the daily influx of reports, and from being viewed by the public for prolonged periods.
==== Badge Permissions ====

Many tickets involve users not being granted badges. Allowing moderation to add badges at their discretion when criteria is reached makes the rectification fair.

==== Site Report Filters ====
Site reports should be filterable by keyword and searchable. The ‘most reported’ filter does not work.

==== Mute Options ====

Additional options for mutes granted by mods can provide more discretion per infraction. A 3 day mute is not a viable enforcement against hate speech and harassment. An option to enforce mutes against comments and reviews, and an option that prevents changing the user’s profile/bots/bot creation.

==== Ban Options ====
Two or more types of ban options should be present: temporary and permanent.

==== Temporary Ban ====
Allows the user to access their account, access their notifications and mod messages, etc. This allows the user to work on rectifying the infraction and welcoming them back into the community of the website while preventing violating content from being visible. This also prevents the rhetoric of ‘bans for no reason’, as the user would be able to log in and view the moderation message stating the reason for the temporary ban.

==== Permanent Ban ====

Allows the user to log in and read moderation messages. Does not allow access to creating bots, personas, lorebooks, chatting, or viewing past chats/bots in the community. The account is visible only by mods and the user (for ban appeal timeframe).

Permanent bans and unbans should require two or more moderators or staff input before being processed. This prevents impulsive (un)bans, accidental (un)bans, witch-hunts, and encourages collaborative review, as content is often subjective. A log should be maintained of unbanned accounts and the reason for each.

===== General Suggestions =====
==== Explicit Content Guidelines ====
State clearly the guidelines users are expected to follow in content creation. An example of a ‘hidden’ guideline is the prohibition of rape in the introduction message. This restriction is not stated in guidelines, but is enforced and results in bot removal.

Users cannot be held to standards that are not transparent. Enforcing hidden or obscure guidelines is not fair or enforceable long term. Transparency lessens the change of re-offense.
==== Accessibility Manager ====
A specified set of tasks or criteria managed by staff to ensure the website remains accessible by screen readers, adaptive technology, and follows best practices for UI and experience.
==== Account Management Portal ====
Account management and history should be viewable by moderation. This allows easy read-only access to view past infractions, moderation actions, etc.
==== Moderation Training ====

Implement shadowing and collaboration-only periods. Moderation is subjective and should be learned by doing. Assigning a moderation mentor to a new hire provides support when viewing heinous material, guidance when making judgements and allows new moderators a chance to use real moderation tools with a safety net for a set period of time.

Clear organization structure. All moderation and staff should know what the company structure is, who to go to for what, who is lead moderation, etc. This prevents developers from being pulled away from their role, or moderators from being unsure on processes. A hierarchy gives transparency and structure.

All moderators should be using an alt account on Janitor, Discord, and Reddit moving forward. Masking moderator user accounts does not prevent or circumvent user harassment. Standard practice when hiring should involve creating a wholly unique email address, Discord account, Reddit account, or Janitor account. These should all share the same name. ‘ModeratorA@gmail.com’ is identifiable as ‘ModeratorA’ on Discord and Janitor, for example.

Volunteers, if any, should be restricted to set hours decided ahead of time. These set hours ideally help fill in times when paid moderation and staff are unavailable.

==== Standardized Language ====

Moderation and staff should be provided messaging for users ahead of published changes. This ensures all users get the same responses from Janitor regarding new content or changes. 
Document and resource creation for staff and moderation should be maintained and created using standard language. Procedure documentation, such as a Best Practices Manual, should be provided to all paid staff and moderation.

==== Developing Norms ====

Regular meetings between developers and moderation (all branches) encourages collaboration and a unified front. Instead of having a channel where mods send messages to devs and wait for answers, a monthly all-hands meeting provides space for development to showcase new features and changes and grants moderation a structured time to ask questions. Critical guidance should still be communicated between development and moderation. Interdepartmental communication should be structured.

==== Conclusion =====

This list was collected from former and current moderators and volunteers and curated into a set of actionable and realistic topics. These are not directives, but the efforts of many to collaborate and propose solutions to many of the issues expressed over the weekend.

Source: https://docs.google.com/document/d/1fNLgqm-7ZPCG6RhQNtZRSmzvsWhTHMTvn8tUtn4nE2Q/edit?usp=sharing