Moderation

Overview

Moderation on Sonata is fully transparent and recorded. Moderators have no power to directly ban users or disable content.

Instead, they can only apply policy violations to content or user profiles. Sonata itself then decides on how to act - either by applying amplification penalties or by suspending or banning the user.

We firmly believe in forgiveness and second chances. Except in the most serious cases, all moderation decisions on Sonata have a time limit. After a maximum of 1 year, all amplification penalties will be fully removed.

All Moderation on Sonata is split into two groups: Prohibited Content and our grey area Amplification Policy.

Prohibited Content covers unacceptable and illegal material. The consequences of this material range from a warning, to total ban, depending on the nature. In all cases, the content is disabled.

Amplification covers socially unacceptable content. The consequences of this material ranges from a short term reduction of amplification level, to a long term one. The content itself is never removed, but it will likely not be shown to other users.

Reports

All content and profiles on Sonata can be reported, like with most platforms. We will try to handle these as quickly as possible.

Appeals

Once a policy violation has been given out, the receiving user may appeal it. This will be addressed and if accepted, the content will be restored, the violation removed and the penalties undone.

For amplification issues, other violations will have their durations re-calculated as they may have been affected by a multiplier that no longer exists.

Sonata's appeal system is among the most transparent of any platform, as all users may choose to publicise the appeal with a button within the appeal page. This will create a static page accessible to anyone with all of the details of the report, with a few personal things removed, making it easy to hold us to our polices and the user to their word.

Automated Moderation

We will conduct some automated checks and lookups to aid our moderation efforts. This will never include the use of non open source LLMs (such as anything by OpenAI), as they would use the data to further their LLM training. If we make use of any LLM, it will be one that is fully open source and serviced by a third party provider that does not record inputs.

Future Plans

Sonata will soon introduce community moderators, users that we have deemed to be sufficiently trustworthy. These moderators will be able to apply policy violations and respond to some reports.

This is possible due to the fully traceable moderation system, it will be easy to spot if a moderator is not acting in good faith.