Comment by lonk11 | Hacker News Reader

Comment by lonk11 | Hacker News Reader

The core problem here is the shared resource - the group. It is a shared channel of information. When you have people with different backgrounds any message on this channel will be "signal" to some and "noise" to others. That causes a conflict over what belongs to the group and what doesn't.

The most common split is how knowledgeable you are on the topic. Experts are your "possums" and beginners are your "otters". If you don't stop the inflow of the beginners, then you get mostly noise for the experts (Eternal September). If you do stop the inflow - you become beginner unfriendly.

The article's suggested solution to regularly split the group tries to lessen the tension on the shared resource by creating more resources. But it's hard - you will lose members as you try to constantly create new groups.

I'd like to suggest a different solution - let users dynamically connect to people who post "signal" and disconnect from people who post "noise". Imagine a group system where when you upvote a post, you connect stronger to other people who upvoted this post (this includes the author of the post). And if you downvote some content you weaken your connection to people who upvoted it. And then how strongly you are connected to others determines how the content is ranked for you. The stronger you are connected to "possums", the more "possumy" content you get. And if you prefer "ottery" content - upvote it and you will see more content from "otters".

This removes the tension because everyone has control over the content they see and they no longer need to fight over the rules of what belongs and what doesn't. This replaces a collective decision with an individual decision.

And guess what, I am building just such system as a hobby project at https://linklonk.com so you can try it out.

I wonder if this is recreating groups without labeling them as such. To a newcomer, such a big group looks like it is composed of many smaller groups. The newcomer then first has to reduce the signal of uninteresting group before finding relevant content (or not and wasting quite a bit of time).

Recreating groups is a hard reset on connections between users. That is, if you migrate from the old group to a new one, you effectively disconnect from all members of the old group and connect to the new.

With LinkLonk you gradually update your connections as you upvoted/downvote content. In a regular group your connection to all other members is a boolean 0 (you are not a member of the group) or 1 (you are a member of the group). On LinkLonk your connection is a float - 0.01 or 0.5 or 3.7 - depending on how useful you found their previous contributions.

As for taking the time to onboard, I agree, this is a problem for LinkLonk - the initial set of recommendations is purely popularity based and there is no topic focus to it. But try upvoting a single item and you will get connected to the other users that upvoted that item. Refresh the recommendations and you will see more content from those users, which should increase the signal-to-noise ratio. Or try submitting a new link and you will get connected to the RSS feed that posted that link.

This is similar to what e.g. the TikTok "for you" algorithm does, right? Interacting with content will make it more likely for you to see similar content in the future.

Cool site!

Yes, the basic idea is similar. The big difference is that the TikTok's algorithm is a black box, while LinkLonk is intentionally transparent.

By transparent I mean that you know what the algorithm will do when you upvote an item. For example, it tells you when press upvote: "You will see more content from 4 users and 3 feeds that recommended this item". And then, when you see a recommendation, it tells you which items connect you to the users that brought you this recommendation (it is shown above every recommendation if you expand "from X users and Y feeds").

My hope is that the users will have more control over the information they get and, as a result, will be more thoughtful about what they upvote/downvote.

Reply on news.ycombinator.com