The comments are broken, and the trolls are winning. Women, people of color, members of other marginalized groups are being pushed out of online spaces through threats and harassment. Some news companies are stopping comments altogether.

We've worked in online communities across many different countries, and we believe in another way. Comments can bring valuable insight, and reader engagement for sites of all sizes. We have to do them better. We think we know how to do that.

We've talked to people who have suffered online death threats, and have formed their own ad hoc communities, using a hodgepodge of tools to attempt to tackle trolls. At the start of this hackathon, we created an online survey, receiving valuable insights about comment platforms from moderators around the world.

As a result, we have built Safety in Numbers. It's a better system for exhausted moderators and for media companies, using a new approach that rewards good behavior, shares the labor of moderation, and prevents trolls from dominating the space.

There are two key innovations in Safety in Numbers. First, the system is designed around a decentralized network of trusted sites. This means that the actions of one moderator improve the system for everyone else - but only for those within the network, because what is seen as abuse on some sites can be normal conversation on another.

This allows like-minded moderators to team up across different media properties, to reward good commenters, squash trolls, and, crucially, support each other. Also, this can be monetized: big media companies can charge smaller ones for access to their moderation network.

The second innovation is algorithmic. We use hidden coefficients to weigh the impact of every commenter's actions. If they're not verified, if they're acting like a bot, if they are flagged for abuse, then any action they take - whether voting a comment down or flagging as spam - has less impact than those of a trusted commenter. We use a series of measures, including specially designed cookies, to track users whether or not they register on the site.

That's our creation: each implementation of Safety in Numbers has its own network and its own rules, backed by ever-changing reputation coefficients for each commenter. This variety of standards, functions and measures makes it much harder for trolls to game the system, and rewards positive long-term engagement across each comment network.

The code that powers Safety in Numbers can be plugged into just about any web platform. We've created a WordPress plugin to prove its functionality, and are excited to see it implemented in a variety of other systems.

We believe that comments need to improve, and that with Safety in Numbers, we might just have a tool to make that happen.

Share this project: