Moderation is a headache for any social network, especially decentralized platforms like Farcaster. Bots and spam push away legitimate users and drown good content. On the other hand, developers expect permissionlessness and decentralisation. Control over recommendation algorithms is the modern equivalent of absolute power, and no one appreciates a monarch.

Farcaster is running an experiment in moderation that is extraordinary in two regards:

• First, they delegate the decision-making power at the channel level (community discussion forums, similar to subreddits), and let any channel owner recommend content their own way.

• Second, the core mechanism is to empower moderators to elevate good content, instead of censoring bad content.

The core mechanism is a binary system where channel moderators or bots flag quality content. Anyone can post in a channel's, but the main feed the users see by default is composed of posts selected by the moderators. The algorithm for good content can be simple rules enforced by a bot like Automod ('poster has a power badge'), complex classification models predicting engagement, human curation, or a combination.

The main benefit of this system is giving developers maximum flexibility. Channel owners or app developers can plug their own moderation/filtering algorithms. Some channels will be human curated, others will use X-like algorithm for engagement, and some might use Reddit-like upvotes. Second, third-party developers can consume that data to inform their own recommendation/moderation engine.

I look forward to seeing many moderation & recommendation experiments. Channels and their interfaces/clients are products with a job to be done, and recommending the right content is crucial. Product support channels (like /neynar or /supercast) might benefit from upvote-based systems, while personal channels (shameless plug for /luc) rely on individual curation. Flexibility allows different channels with varied content to use suitable content surfacing mechanisms.

Philosophically, I find the principle of elevating over censoring to be more appealing. No one envies the hired moderators who are subjected to violent and graphic content on X, Facebook, and other social platforms, like the characters in 'A Clockwork Orange.'

As the cost of content creation trends to zero with generative models, personal and positive curation increases in value. Farcaster channels have a great opportunity to be the place where the best content is created and surfaced.

Moderation is a headache for any social network, especially decentralized platforms like Farcaster. Bots and spam push away legitimate users and drown good content. On the other hand, developers expect permissionlessness and decentralisation. Control over recommendation algorithms is the modern equivalent of absolute power, and no one appreciates a monarch.

Farcaster is running an experiment in moderation that is extraordinary in two regards:

• First, they delegate the decision-making power at the channel level (community discussion forums, similar to subreddits), and let any channel owner recommend content their own way.

• Second, the core mechanism is to empower moderators to elevate good content, instead of censoring bad content.

The core mechanism is a binary system where channel moderators or bots flag quality content. Anyone can post in a channel's, but the main feed the users see by default is composed of posts selected by the moderators. The algorithm for good content can be simple rules enforced by a bot like Automod ('poster has a power badge'), complex classification models predicting engagement, human curation, or a combination.

The main benefit of this system is giving developers maximum flexibility. Channel owners or app developers can plug their own moderation/filtering algorithms. Some channels will be human curated, others will use X-like algorithm for engagement, and some might use Reddit-like upvotes. Second, third-party developers can consume that data to inform their own recommendation/moderation engine.

I look forward to seeing many moderation & recommendation experiments. Channels and their interfaces/clients are products with a job to be done, and recommending the right content is crucial. Product support channels (like /neynar or /supercast) might benefit from upvote-based systems, while personal channels (shameless plug for /luc) rely on individual curation. Flexibility allows different channels with varied content to use suitable content surfacing mechanisms.

Philosophically, I find the principle of elevating over censoring to be more appealing. No one envies the hired moderators who are subjected to violent and graphic content on X, Facebook, and other social platforms, like the characters in 'A Clockwork Orange.'

As the cost of content creation trends to zero with generative models, personal and positive curation increases in value. Farcaster channels have a great opportunity to be the place where the best content is created and surfaced.