June 8, 2021

Moderation Clarification to the Community

In this update, we address questions we’ve received about moderation policies and privacy, and introduce our new Content Policy with specific guidelines about what content is allowed on AI Dungeon.

As Latitude and the AI Dungeon community grow, we want to take the opportunity to be more transparent by creating new, clearer Community Guidelines and a separate Content Policy. The new format should help users understand what content is allowed on AI Dungeon, and what content is allowed in published or multiplayer adventures. We have also standardized our moderation process with new internal tooling to support faster, clearer, and more effective responses to community content reports and other flags. In this update, we address questions we’ve received about moderation policies and privacy, and introduce our new Content Policy with specific guidelines about what content is allowed on AI Dungeon.

Moderation Process

Latitude’s content moderation process starts with a flag.

Users can flag AI output with the

Screenshot of the Flag button from the AI Dungeon UI.

button while playing.  Users can also report other users’ published content or contributions to multiplayer adventures with the Report option on scenarios, posts, comments, and adventures. Recently, we also added a filter that automatically flags some content that may violate our policies. None of these flags trigger a permanent suspension without human review.

If the filter flags a user's content, the system may temporarily suspend the user’s account to prevent abusive behavior. Our human moderators then review the flagged content to determine whether a permanent suspension is warranted or if the content was flagged incorrectly. When moderators identify incorrect flags, the temporary suspension will be lifted immediately. We do not automatically suspend users for content flagged by other users because of the risk of brigading (where many people falsely report content to harass the poster), but it does enter the queue for review by our moderators.

Whether users or the automated system flag content, moderators who are trained Latitude employees or, in some cases, members of Latitude’s legal team, review flagged content to determine whether further action is necessary, including lifting automatic temporary suspensions before the suspension expires if a flag was incorrect. Latitude does not share flagged content to non-expert reviewers (such as crowdsourcing platforms) for moderation purposes.

When a moderator reviews a flagged action from an unpublished adventure, they initially see only the one prompt submitted to the AI that triggered the flag. In some cases, they may need more context, and can view additional actions or AI responses from that adventure as needed. This lets us ensure that we’re not taking a single sentence out-of-context, and aids in our context-based approach to moderation. Our reviews of flagged content are logged, including the reviewer and the decision they made, so Latitude knows which employee or legal team member took a particular moderation action.

Visual representation of the moderation flow described before this; information in the chart is a subset of the preceding text.

Content Policy & Community Guidelines

Previously, you had to refer to both the Community Guidelines and the Terms of Service for information about what content was allowed on which parts of AI Dungeon. That made finding specifics of whether a storyline would be allowed hard, and some users were unsure of whether their adventures would cause their account to be banned.

We heard that feedback, and have broken our Content Policy out into a separate document from the Terms of Service and previous Community Guidelines. We may still suspend accounts for violations of the Terms of Service that aren’t content-related, but all content-related moderation actions will now refer to the Content Policy. Additionally, our Community Guidelines are now focused on interactions between players.

Our new Content Policy and Community Guidelines are included at the end of this post for your convenience, but you can also find them at Community Guidelines and Content Policy respectively. Thank you for your patience while we adjust our policies to account for our growth, and for playing AI Dungeon.

Community Guidelines

Latitude created AI Dungeon to encourage people to share games, support others, and allow players and game creators to connect directly.

We want everyone to have a good experience playing AI Dungeon, and the vast majority of users interact with others in a responsible and safe manner. If you do experience an interaction with another user that may violate our guidelines or other policies, please report it to us at [email protected] If we determine the conduct reported violates our policies, we may take Enforcement actions (see Enforcement in our Content Policy for more information) based on the severity of the violation.

Our guidelines below complement our Content Policy, and govern our expectations on what is and isn’t allowed in how people participate both online and offline in our communities.

One: Be kind and respectful to others

It is normal to disagree with others, don’t harass others. Harassment includes encouraging violence, using hateful language, and threatening behavior. People should feel safe to participate in healthy discussions.

Two: Respect the privacy and boundaries of others

If a user blocks you, respect that boundary and do not circumvent it (such as by creating new accounts hiding your identity). You also may not share or threaten to share someone’s personally identifiable information or private content without permission. See Harassment in our Content Policy for more information.

Three: Help facilitate predictable experiences

Help others set their expectations by using predictable and accurate tags for your content, and use the NSFW flag when engaging around permitted mature and sexual content.

Four: Keep the platform, yourself, and other people safe

We ask that you respect our platform and other users by not engaging in malicious activity, such as selling or facilitating sales of prohibited or illegal goods and services; creating or distributing illegal content like sexual content involving minors, malware, or viruses; hacking others; and coordinating or assisting others in making false reports.

Relatedly, while we do not allow players to impersonate individuals or organizations with intent to cause harm, we don’t recommend sharing your real name or other personal information. It is not okay to create multiple accounts to influence others, such as inflating your follower count or upvoting your own content.

Additionally, we ask users not to seek therapy, medical, or other professional advice on our platform. If you are experiencing distress and are considering self-harm, help is available through the US National Suicide Prevention Hotline at 1-800-273-8255. We prohibit activity that may jeopardize the life and safety of yourself or others offline.

Content Policy

The following policies outline our expectations and Latitude’s approach to user provided content. You are expected to follow these policies in all of your activities on AI Dungeon, including but not limited to:

  • Shared content (whether published or not)
  • Unpublished content
  • Multiplayer content

Violating Latitude’s policies can result in a formal warning, temporary account suspension, or permanent account suspension and closure based on the severity and type of violation. We will consider a user’s existing pattern of behavior, previous history of violations, severity of the content, and context when evaluating content to inform any resulting enforcement actions. Latitude has internal policies, procedures, and systems in place to apply these policies in a fair and consistent manner.

For more information about the information Latitude collects to prevent violations of our policies, please see our Privacy Policy. Latitude reserves the right to remove content, close accounts, or take additional action at our sole discretion, with or without advance warning.

Our Community Guidelines include additional information specific to community interaction, both online and offline.

Violence

As a gaming platform, Latitude takes a context-based approach to violence in games. The following are examples of permitted content:

  • Fictional evil characters and/or villains with dialogue that expresses evil intent
  • Story violence against enemies
  • Fictional scenarios involving violence, possibly including but not limited to weapons, magic, war, assassination, spy activity, and covert ops.

Latitude does not allow content that incites hatred or violence offline, including content encouraging, supporting, glorifying, or trivializing:

  • Terrorism
  • Violent extremist actors and/or acts
  • Hate groups and/or hate group membership
  • Terrorist or extremist propaganda
  • Gore and other obscene content
  • Violence against others (see Harassment)

Threatening Behavior

Latitude has zero tolerance for threatening behavior, including threatening the life, safety, or well-being of others online or offline. We permanently suspend accounts associated with such content.

Mental Health & Self-Destructive Behavior

We do not permit use of our game as a therapy alternative, to seek medical or other professional advice.  If you are experiencing distress and are considering self harm, help is available through the US National Suicide Prevention Hotline at 1-800-273-8255.

We also prohibit content that explicitly encourages, supports, glorifies, or trivializes activity that may jeopardize the life and immediate safety of yourself or others offline. This may include, but is not limited to:

  • Expressing suicidal intent or threats
  • Physical injury to self
  • Threatening behavior meant to cause harm to yourself

Harassment

Latitude does not tolerate harassment of others online or offline harassment related to use of our Services. This may include, but is not limited to:

  • Targeted harassment or personal attacks
  • Inciting or promoting violence or abuse against others
  • Stalking or ignoring physical boundaries
  • Unauthorized sharing of personal information including:
  • Personally identifiable information (e.g. real name, location, or medical records)
  • Private content without permission
  • Sexual harassment including:
  • Unwelcome sexual advances or solicitations
  • Sexual objectification
  • Sharing or threatening to share private sexual content without permission of the person depicted

Examples of content permitted in context include:

  • A character, Robin Hood, levies personal attacks and incites abuse against another character, the Sheriff of Nottingham
  • In a game about a woman becoming an esteemed lawyer, she experiences discrimination and an unwelcome sexual advance amidst a legal case

Hateful Conduct

Latitude does not tolerate content intended to vilify, encourage, incite, or promote hate, violence against, or threatening other people based on actual or perceived race, color, religion, religious creed, national origin, ancestry, citizenship, physical or mental disability, medical condition, genetic information, marital status, sex, gender (including gender identity and gender expression), age, sexual orientation, veteran and/or military status, or domestic violence victim status. Hateful conduct examples include, but are not limited to:

  • Using hateful language in your username or profile bio
  • Accounts dedicated to inciting violence against people of color or other groups
  • Games centered around wishing harm to vulnerable people

The following are examples of content permitted by our context-based approach:

  • A game where a disabled character describes themself in terms that may otherwise be disallowed
  • A game describing the nuances of reclaiming slurs in the LGBTQ community

Impersonation

Users are not permitted to impersonate individuals or organizations affiliated with Latitude on or off of Latitude’s platforms. Accounts and/or content that mislead or misrepresent oneself by impersonating individuals or organizations with the intent to cause harm, injury, loss, or to defraud others are also not permitted. Examples of this include:

  • Impersonating a Latitude employee or other users
  • Misrepresenting and claiming affiliation to a partner organization

Mature & Sexual Content

We prohibit content that depicts, encourages, promotes, or refers to:

  • Any sexual content involving individuals under 18 years old fictional or otherwise (see Sexual content involving minors for more information)
  • Nonconsensual pornography (any sexually explicit material featuring an individual who has not expressed fully informed consent to such material being posted and shared, or who has since revoked their initial consent)
  • Necrophilia
  • Bestiality
  • Sex trafficking
  • Content that is obscene or illegal under applicable local law

We may permit some mature or sexual content with a context-based approach, including, but not limited to:

  • Consensual adult content, including furry content
  • Violent content (see Violence for more information)
  • Swearing and adult language in general contexts

Sexual content involving minors

Latitude applies a zero-tolerance policy for sexual or suggestive content involving minors, including child sexual abuse material and imagery; fantasy content (e.g. “loli”) that depicts, encourages, or promotes the sexualization of minors or those who appear to be minors; or child sexual exploitation, with a permanent account suspension for all accounts associated with such content.

Labeling

We encourage you to use predictable and accurate labels when tagging your content. This will enable users to find the content they’d like to see more quickly. For content that is permitted Mature and Sexual Content, applying the NSFW flag is required. Spamming of irrelevant or inappropriate labels is not permitted.

Enforcement action may include, but is not limited to, a written warning or temporary suspension, adjusting the tags to correct inaccuracies or omission of the “nsfw” tag, and permanent application of the NSFW flag on all content from the violating account.

Malicious Conduct

Latitude prohibits any content or activity that may be used to mislead, harm, or threaten public safety. This includes, but is not limited to:

  • Harmful misinformation (content that is misleading or contains false information)
  • Example: a game that uses biased data with a manipulated narrative to express anti-vaccination sentiment
  • Disinformation (deliberate and intentional spread of misinformation)
  • Example: purposefully generating and spreading a scenario related to a major event to incite violence
  • False or misleading political marketing content to persuade or change the minds of an audience
  • Example: generating false or misleading information about politics and elections to deceive or harm voters, including offline or off-platform

We additionally prohibit the following types of content for their ability to violate our policies, influence, or harm:

  • Advertisement content generation or distribution
  • Scams
  • Spreading malware or viruses
  • Artificially inflating metrics or game statistics, like number of upvotes or followers
  • Bypassing Latitude systems to generate content that violates our policies (see also the “Restrictions On Your Use Of The Services.” section of our Terms of Service)

Enforcement

If an account is found to be in violation of our policies, we will enforce our policies. Depending on the nature and severity of the violation, enforcement may be without prior notice to you. This includes, but is not limited to:

  • Issuing a written warning
  • Temporary or permanent account suspension
  • Removing content that violates our policies
  • Permanently applying a NSFW setting on the violating user’s content

Attempts to circumvent an account suspension, temporary or permanent, are prohibited. Users may petition for the removal of their suspension or wait until the end of the suspension expiry period described in the notification of your account suspension.