Comment by DanielBMarkham | Hacker News Reader

Comment by DanielBMarkham | Hacker News Reader

Parent Full thread

Something about this bothers me. Can't put my finger on it.

I think the assumption is that there are articles so bad that it should be obvious that they are off-topic. In the past, I've submitted things on the border. I imagine many others have too. I've always counted on the groupthink to correct any errors in submission I might make. This seems to assume that there are hard guidelines. Watching the board over the years, I'm not sure that assumption is accurate.

Put differently, if you had a cache of really bad articles, shouldn't we see them? That way we'd know not to submit. But if you already know they're bad, then what's the point of voting or flagging?

Perhaps I'm just mentally adrift here. Honeypots make sense to me when we are talking about boolean things: a website visitor is either harmful or not. An email sender is either a spammer or not. But I'm not sure at all that this concept applies to something like an essay. Seems like if it would, you could just use the flagging behavior mentioned to rank the articles and dump everybody else's votes. Right? This is like verifying the voting behavior by setting up some completely different system to rank quality detectors. But if you could rank quality detectors, why keep the old system? And if not, how would you separate which parts of which system are useful and which are not?

I think it is more and more becoming clear that google's approach is the best way: your ideal first page has to be personalized (when you are looged in, Google personalizes your results with location, previous searches, +1s, and even who you follow on twitter).

Trying to be everything to everybody means there will be people left with sub-optimal results.

For those downvoting: Please share your thoughts. If you disagree, I'd love to know why.

It's not you; it's that pg has said many times (and I agree) that fragmenting the front page is a bad idea. We must all see the same front page to judge the same quality.

Interesting. However, many people already filter the first page, let it be by points, or by the twitter accounts:

http://twitter.com/#!/newsyc20

http://twitter.com/#!/newsyc50

http://twitter.com/#!/newsyc100

http://twitter.com/#!/newsyc150

And from each filter, people auto-select things that interest them. Sometimes I only see a story when it is retweeted to me.

And this you can't prevent. It lies on the fact that different people have different definitions of what "quality" means.

Which is the core problem highlighted by linked blog post.

The front-page of HN is a filtered list in and of itself. And I for one actually like the result.

Some people would like to see a different filter and thus create one, because they can. That is indeed not something you can, or should want to, prevent.

But the fact that some people create their own filters is not a motivation to not tweak the HN front-page filter in such a way that the front-page matches the intended goal (pg's goal in this case, presumably adopted by the majority of HN readers, more-or-less codified in the guidelines) as closely as possible.

"The core problem ... lies on the fact that different people have different definitions of what "quality" means."

The user is not the problem. The user is the solution. http://news.ycombinator.com/item?id=3170840

This becomes suboptimal on a news site. I don't want to just see content that I know and is interesting to me, I want to read things I would have never have a chance of coming across otherwise. Of course, it is also impossible to read everything. This is why the voting system works so well. If many users find an article interesting, chances are that it will be interesting to everyone. This dynamic is increased substantially with a specific group like Hacker News. This is why I don't like personalised pages, they are only really feasible for things like google+ and certain types of searches.

I agree with this. However I don't think that existing communities can easily adapt to this well. Instead I think it has to be something that is part of the initial design of the platform so that it's tightly integrated into the service. You could go further than Google and actually deeply integrate a social graph with feed weighting based on who you follow and what they vote on as well as your previous voting record and maybe some shortcut to topic, such as tags.

Why aren't we taking the bayesian approach?

Can you be more explicit?

I'll bet he want to complificate the formula to look something more like

"P(H|Q)=P(Q|H)P(H)/P(Q)" where

H = "User upvoted a honeypot article" and

Q = "User's votes are not a good signal for article quality"

And a similar adjustment for flagging.

The concept of a 'honeypot' article is a little ridiculous.

The best you can hope for is "this preferred population of people overwhelmingly dislikes this article". The problem at hand is that you want foster the interests and tastes for that one group of people.

So instead of crusading against lame newbs with a labour intensive system of 'silly articles' (who picks what is an article that ought to be downvoted?), you could compare their weighted voting history against your population of 'good users', yadda yadda yadda.

Bayesian approach could be nice uf it can be used to increase diversity.

Sounds like a machine learning problem. Showing people the cache of really bad articles would lead to overfitting.

Making each user's vote a vector with the value of the honeypot formula determining its strength, and the vote total for any article a floor function of its total votes would be pretty cool, but might be too computationally expensive--certainly more so than the user-categorization approach.

I don't see any problem. The article is talking about ranking up voters and flaggers, not submitters, so you could continue with your submissions and not be penalized.

Reply on news.ycombinator.com