Show HN: My thoughts on solving Eternal September | Hacker News Reader

Show HN: My thoughts on solving Eternal September

2 pointspwatsonwailes8 comments

Over the past couple of months, I've been considering the problem of the decline in comments on HN, against sites where I've seen the quality hold over time. As a result of this, I've come to three conclusions:

1. Community size (that a user can see & engage with) must be artificially limited (although total community size doesn't) 2. A weighted voting process needs to be in place 3. To scale, the controls must be algorithm-based, not moderator-based

Therefore I've come up with the following system that I'm going to be building and trialling on a site in the near future:

1. When a user joins, they're initially put into the city population. Here they are shown the firehose - an unfiltered news stream, with all users present

2. Each time a user interacts, either by adding a comment, posting an item or clicking a link, the action is recorded

3. Over time, the accrued weight of this will ensure that what they're shown is limited by the types of content they've historically interacted with, the users who've submitted that content, and the users they've engaged with in commenting

4. To ensure freshness/adaptability, a certain amount of the hose will always be shown, and the selection will have a time component, so that they can view different content and thus the selections can change as the user changes

5. Users will only be exposed to other users who loosely fit the profile that they've selected in to, with a maximum of around 150. Users will be able to follow other users, thus ensuring that these people are never swapped out of their engagement circle, and all content or comments that they submit are presented

6. To solve the obvious issue of thousands of people following single users, a power user will see content from their own followed list first, then other data below

Looking at the content side:

1. Each item would be tagged according by key terms in the content & the user who submitted the item

2. Each submission & comment would allow voting, thus allowing for segregation between casual browsing and active engagement, to allow for more refined user-level filtering

Thoughts/suggestions appreciated.

1. Community size must be artificially limited

This might end up being a good way to kill the community. If something stops growing, it typically starts dying. A better goal is to try to find ways to problem-solve such that it supports healthy growth rather than willy-nilly growth. Supporting healthy growth may also have a side effect of slowing growth, which can be good, but artificially limiting size is likely to be a very bad idea.

2. A weighted voting process needs to be in place

Why?

3. To scale, the controls must be algorithm-based, not moderator-based

Rules-based crap that isn't adequately supported by the wisdom of good human judgment tends to lead to the sort of bureaucratic nightmare that people whine about concerning the government (et al). So I doubt it.

1. Should have been clearer on point 1 - it's not limiting the size of the community, but the size that any individual user can see. I'll edit this

2. To allow users to self-segment themselves to keep the content they see tailored to their interests

3. Of course there needs to be a human level, but ideally this would be at an administration, not a workings level

2. To allow users to self-segment themselves to keep the content they see tailored to their interests

Please elaborate. This sounds interesting and has potential.

So my theory is, you can tailor content based around 3 things:

1. The submissions that a user views (and adding greater weight to those they vote for)

2. Comments that they respond to (again incrementing the weight if they vote), and

3. Users who's content (in the form of comments/submissions) they react to

So if I consistently view content submitted by a user, and respond to their comments a lot, I'd be shown more of their content. Also, if I always click on titles containing HTML, CSS, PHP and jQuery, I'd see more links with those sorts of things in them.

Conversely, as a result of not clicking on things, over time I'll see less of them.

By having say, a six month drop-off period, this has the dual effects of increasing the likelihood of me seeing content relevant to me, and this content staying current as my interests change.

Also, by showing the user similar users (based on engagement habits) over time, it makes it more likely that they'll:

1. Engage more frequently, and

2. Have better discussions, as they get to better know the members of the community that they regularly see

Old data (and I can't really tell you the source exactly -- possibly internal data from when I moderated something and the owner had done some measures combined with some external stats), FWIW:

As I understand it, about 20% of members of a forum will be "active" posters and another 10% will be "occasional" posters. According to data in "The Tipping Point", the average/typical human mind can follow the interactions of about 150 people in a group setting. In my experience, the above combined stats results in a community of up to around 750 members will result in around 150 active members and works just fine. Above that and things begin breaking down socially. As far as I know, the only solution humans have come up with so far to counteract/cope with communities larger than 150 is formality of social interaction. I think the transition from warm cozy community of about 150 active members to the need for something more formal to keep coping as it gets bigger is the growing pain that most communities seem to trip over the worst.

Maybe you can incorporate some of that info into your thinking.

Peace.

Cheers! Have a great day!

What you refer to as the firehose --- what people are exposed to in your scheme when they first register --- is also what is shown to people who have not registered. They are people who we would either like to register and join in (if they would improve the site) or not register (if they would detract from it). We want them to either be attracted or repelled by the mix of links and discussion on offer.

If all that we see is personalised to us, we lose the ability to shape what they see, and we lose the ability to attract people who will contribute to the community we want to see.

I guess you could use the most followed users to create a default-filtered firehose, but whatever happens there's going to need to be some form of default setting, where people have to start filtering to refine results (even if it's just by tag a-la Reddit).

Any other ideas about how to solve that?

Reply on news.ycombinator.com