moderation/README.md

# Moderation guidelines

## Status

This document is a working draft and incomplete.

Missing aspects:
- spam
- automated posts
- high-frequency posts
- …

## Non‐normative summary

We wish to provide an environment in which queer and neurodivergent people can
be their authentic selves. In addition to communication itself, we have various
tools for restricting communication to help us achieve that goal.

We need however be mindful that every action we take may not only protect from
harm, but can also be harmful by itself. We should therefore in general respect
people’s autonomy in deciding with whom they wish to interact and what kind of
content they like to see. We should only limit that autonomy to a degree that
is necessary and effective to achieve our goals.

Distribution of content that is strictly illegal under the laws of the
Federal Repulic of Germany, whether we agree with them or not, obviously
undermines our goals because it may result in termination of this instance.

Communication may be harmful to others, both to users of this instance as well
as on other instances. Harmfulness can be difficult to assess and quantify
objectively and always requires consideration of the larger context.
Communication content can be annoying or distressing to people, or it might
cause them to take harmful action.

Our moderation actions should be proportional to the risk of harm resulting
from the communication we moderate. We should attempt cooperative measures
(discussing an issue with the originators of a communication or the moderators
of another instance) before we forcefully restrict communication.

We should also note that we have the ability to apply restrictions selectively
both in terms of the communication parties or content as well as the degree of
restriction.

Restrictions can be applied based on local user identity, remote instance name,
hash tag or content keywords. They may be applied to messages themselves or
media attachments. And the degree of restriction can vary from a reduction in
visibility to complete termination.

### Local users

Local users are generally moderated through the administrative web interface
of this instance and available moderation actions include

- unlisting (removal from federated timeline), `mrf_tag:force-unlisted`,
- sandboxing (removal from public timelines), `mrf_tag:sandbox`,
- marking media attachments as sensitive, `mrf_tag:media-force-nsfw`,
- stripping media attachments, `mrf_tag:media-strip`,
- account deactivation, and
- account deletion.

### Instances

Instances are moderated through this Git repository using the following
settings:

- `activities`:
  - `unlist`: Remove activities from federated timeline,
  - `restrict`: force activities to be visible to followers only, or
  - `reject`: reject all activities except deletes.
- `media`:
  - `mark`: Mark media attachments as sensitive, or
  - `strip`: strip all media attachments.

### Hashtags

Hashtags can be matched by case‐insensitive exact match with the following
flags:

- `sensitive`: Mark tagged activities as sensitive, and
- `unlisted`: remove tagged activities from federated timeline.

### Key words

Key words are matched by regular expression with the following options for
moderation:

- `unlist`: Remove matching activities from federated timelines, or
- `reject`: reject matching activities altogether.

## Definitions

The key words _shall_, _shall not_, _should_, _should not_, _may_ are to be interpreted as described in
[RFC 2119](https://www.rfc-editor.org/rfc/rfc2119.html).

- _shall_: strict adherence, no deviation
- _should_: recommendation, deviation with good reason
- _may_: suggestion, optional, but good manners

### Representation

The term _representation_ applies to both textual (something being described in text) and pictoral
(something shown in a still image or video) representation. It applies whether or not the represented
idea is real or imaginary.

## Visibility guidelines

- _unrestricted_
  - Activities _may_ be visible in public timelines.
  - Media attachments _may_ not be marked as sensitive.
- _restricted_
  - Activities _shall not_ be visible in public timelines if unlabeled.
  - Activities _should_ be labelled with a content warning.
  - Activities _may_ be visible in public timelines if labelled with a content warning.
  - Media attachments _shall_ be marked as sensitive.
- _prohibited_
  - Activities _shall_ be rejected.
  - Media attachments _shall_ be stripped.

### Personal integrity

Many individuals experience discomfort or distress when confronted with the compromise of a person’s integrity.

Personal integrity includes:
- Physical (bodily) integrity
- Mental (emotional) integrity

**Representation of:**
- compromise of personal integrity _should_ be restricted.
  - invasive medical procedures
  - mental health
- non-consensual compromise of personal integrity _shall_ be restricted.
  - physical violence
  - psychological abuse

**Practical application:**
- Media depicting physical violence _shall_ be marked as sensitive and labelled with `violence`.
- Posts about mental health issues _should_ be labelled with `mh`.

### Nudity & sexuality

Nudity and sexuality are considered private to varying degrees in most cultures. Some individuals have
experienced trauma related to sexuality and find confrontation with nudity or sexuality distressing.
Some representations of sexuality are illegal.

**Representation of:**
- any penis, vulva or anus _should_ be restricted.
  - body parts commonly used to experience sexual pleasure
- bare anthromorphic torso _may_ be restricted.
  - unclothed breasts, abdomen or buttocks
- non‐sexual nudity _may_ be unrestricted.
  - breast-feeding of children
  - non‐sexual nudity in medical settings
  - non‐sexual nudity in culturally accepted settings
- erotic or sexual nudity _shall_ be restricted.
- erotic or sexual nudity involving humans or anthropomorphic characters below the age of 18 _shall_ be prohibited.
- sexuality _should_ be restricted.
- sexuality involving humans or anthropomorphic chararcters below the age of 16 _shall_ be prohibited.
  - edge case: person describing their own experience of sexual abuse

**Practical application:**
- Pictures with nudity _should_ generally be marked as sensitive and labelled with `nudity`.
- Sexually suggestive pictures without nudity _should_ be marked as sensitive and labelled with `lewd` or `suggestive`.
- Explicit descriptions or depictions of sexual activity _shall_ be marked as sensitive and labelled with `lewd`.

### Bodily emissions

Emissions from a being’s body tend to be considered private in most cultures. Many individuals experience
discomfort when confronted with them.

**Representation of:**
- bleeding or blood _should_ be restricted.
- micturition or urine _should_ be restricted.
- defecation or feces _should_ be restricted.
- ejaculation or ejaculate _shall_ be restricted.

**Practical application:**
- Descriptions or depictions of blood _should_ be marked as sensitive and labelled with `blood`.
- Pictures of dog excrement _should_ be marked as sensitive and labelled with `scat`.

### Existential threats

Confrontation with existential threats may be important, but can be highly distressing.

**Representation of:**
- existential threats to any population _should_ be restricted.

**Practical application:**
- Discussions of war or climate change _should_ be labelled with appropriate content warnings.

### Politics

Discussion of politics is very important, but can be exhausting. Some political ideologies are fundamentally
incompatible with our values.

**Practical application:**
- Discussion of politics _may_ be labelled with content warnings (`pol`, `depol`, `uspol` etc.).

### Food

Food and its consumption are not considered private in most cultures, but some individuals have eating disorders
and may experience distress when confronted with food.

**Representation of:**
- eating or food _may_ be restricted.

**Practical application:**
- Pictures of meals _may_ be marked as sensitive and labelled with `food`.

### Eye contact

Eye contact may be experienced as uncomfortable by some people. May need some research.

**Representation of:**
- eyes focused on the viewer _may_ be restricted.

**Practical application:**
- Selfies with eyes focused on the camera _may_ be marked as sensitive and labelled with `ec`.

## Escalation strategy

The moderation action should be proportional to the harm potential of the
communication. Occasional mild infractions should be solved through cooperation
while persistent or serious violations may require forceful action.

Failure to remedy an issue or repeated violation should be met with gradual
escalation of measures.

(incomplete)

### Prohibited content

Content that is obviously illegal or very harmful is _prohibited_.

- _Any_ such content _shall_ be deleted locally immediately.
- Local user accounts distributing _any_ such content _shall_ be restricted
  immediately by sandboxing or deactivation, they _should_ however not be
  deleted without further investigation.
- Local user accounts _primarily_ distributing such content _shall_ be deleted
  immediately.
- Remote instances distributing _any_ such content _may_ be restricted
  immediately by `restrict`ing their activities and `mark`ing or `strip`ping
  their attachments.
- Remote instances failing to contain the distribution of _any_ such content
  within an adequate time period _should_ at least be restricted by
  `restrict`ing their activities and `strip`ping their attachments. They _may_
  also be defederated by `reject`ing their activities.
- Remote instances _primarily_ distributing such content _shall_ be defederated
  immediately by `reject`ing their activities and `strip`ping their
  attachments.

### Restricted content

- _Any_ such content _may_ be removed from the public and federated timelines
  if not adequately labelled.
- _Any_ such content _may_ be forcibly labelled with appropriate content
  warnings.
- Local user accounts distributing _any_ such content without adequate labels
  _should_ be contacted for cooperative resolution.
- Remote user accounts distributing _any_ such content without adequate labels
  _may_ be contacted for cooperative resolution.
- Local user accounts _primarily_ distributing such content without adequate
  labels _should_ be contacted for discussion.

(incomplete)

### Cooperative communication

- Try to be friendly and respectful in your communication.
- Describe your role.
- Clearly describe the offending behaviour and explain the reason for it being
  considered offensive.
- Note whether improvement of the behaviour is a suggestion, a recommendation
  or mandatory.
- Mention the possible consequences if the behaviour is not improved.

#### Examples

> “Hey, I am a moderator of this instance. My colleagues and I noticed that you
> have been posting a lot on current political events. While we have no strict
> rules about it, we feel that an excess of such content is very exhausting to
> our users. We therefore suggest that you consider labelling such posts as
> `??pol`, so that users can skip over them if they are not interested, or
> publishing them as _unlisted_, so that they are not visible on the public
> timelines.
> 
> We trust in your ability to be considerate to others and don’t believe that
> any further action is required on our part.”

> “Hey, I am part of the moderatorion team of this instance. We have received
> complaints from other users about sexually suggestive posts from your
> account. While not considered inappropriate per se, we recommend that such
> posts be labelled as `suggestive` or `lewd` and media marked as _sensitive_.
> 
> Please be reminded that if you continue to post such content without
> marking it appropriately, we may decide to remove your posts from the public
> timelines and mark all your media attachments as _sensitive_.”

> “Hey, I am contacting you on behalf of this instances’s moderation team. You
> have been posting sexually explicit images without marking them as sensitive.
> This is however mandatory and if you violate this requirement again, we will
> restrict your posts to followers only and forcibly mark all your media
> attachments as _sensitive_.”

> “Hi, I am writing to you on behalf of this instance’s moderation team. We
> have noticed that you have been mentioning … despite their clearly and
> repeatedly stated request not to be involved in the discussion any longer.
> We ask that you disengage and honour their wish. If you fail to do so, we
> will have to suspend your account.”