The importance of upvoting | Hacker News Reader

The importance of upvoting

julian.com

181 pointsjulianshapiro62 comments

Consider two users: Bob browses HN and upvotes about 1000 articles a month. Jill, on the other hand, upvotes only articles she considers to be exceptional and upvotes about 1 article a month.

Should Bob have 1000 times more influence than Jill? Personally, I'd like to see what was so special about that one article Jill voted for, but a naive upvote mechanic considers it just a single vote, and is not noteworthy when compared with Bob's thousands upon thousands of votes. It would be nice is Jill's single vote was equal to 1000 votes from Bob.

Do any systems like this exist? It sounds good in theory, but I wonder how it would stand up to abuse?

Any ideas for ways it could be abused?

It should also consider the time spent on the site. If Jill only spend 10 minutes a day on HN, then her 1 upvote would be worth less. But BoB, upvoting 1000 articles, if he only spend the same 10 min day, then his upvote should be worth less on average.

If Bob is upvoting good articles his votes should count.

The problem on HN isn't that many people upvote, but that some people upvote bad submissions.

It might be useful to look at people who regularly upvote bad submissions and down-weight their votes.

That's how digg worked, it did not work well for them.

Most of the time I don't upvote because the link isn't good enough. It's very rare that I see a link I think is really worth people's time - maybe once or twice a week.

I could lower my bar, but the real problem is the binary nature of the vote. There's content that I really want more of, and then there's content that's better than average. If I vote both with my single vote, I'm not delivering the right level of information. It's almost like I want a supervote that can only be applied once a day, or once a week, that's worth more than one normal vote.

I just tried digging for examples of supervotes in the wild. The idea is intriguing.

Couldn't find an aggregator that does it. Interestingly, however, it's a common voting tactic on reality competition TV (e.g. The Voice, American Idol) [1].

Does anyone have examples of supervotes in action for online communities?

[1] https://www.google.com/webhp?sourceid=chrome-instant&ion=1&e...

There's a Discourse-type thing (it's not Discourse) that gives you upto 3 votes on any one topic, and I think a daily maximum, that perhaps increases with user trust - but I don't remember the name.

I see it from time to time used as a forum or feature request tracker.

It's called uservoice: https://feedback.uservoice.com/forums/1-product-management/s...

I would love to see this concept applied to news aggregators

Does tinder's super like count? Obviously a totally different domain. I have mixed feelings about it.

Reddit gold is kind of a super vote. It doesn't actually carry any weight but it does draw attention to the post.

I believe the author is unaware of the 1% rule on internet culture and argues that people shouldn't follow it. Making an impassioned plea isn't going to do much. The internet is full of people that don't want to interact and only consume . https://en.m.wikipedia.org/wiki/1%25_rule_(Internet_culture)

Thanks for pointing this out to me. I had in fact forgotten about this rule.

I'd like to prod that rule, though, and see if being vocal about the importance of upvoting can shift the participation rate on a per-community basis. My hunch is that it can, and Algolia can help us find out over time.

Isn't the more significant question whether the 1% rule actually has to remain valid forever? Is it a law of nature or is it more learned behaviour? Why is there no 1% rule for elections? Because people have been taught that it is important to vote. People on the Internet have not been taught the same thing. The author suggests that this should change. One blog post of course won't do that. But you gotta start somewhere.

I think there is something like a 1% rule for elections.

small group runs for office

larger group helps run the campaign (volunteers, etc)

larger group donates

larger group votes

even larger group doesn't vote at all

As the amount of effort to participate decreases, the amount of people involved rises.

I don't know, this seems a bit like taunting Goodhart's law? [1] I'm thankful that voting on Hacker News works as well as it does; maybe we shouldn't mess with it.

The Hacker News moderators are doing what they can to provide a quality website. To the extent your voting helps them achieve their goals, it's useful. On the other hand, encouraging more noise voters won't improve things; it will mean that quality submissions need more votes to break through the clutter, and the moderators will have to look for alternate ranking signals.

Still, it probably wouldn't hurt to look at the 'new' page more often.

[1] https://en.wikipedia.org/wiki/Goodhart's_law

I think this is a misidentification of the true problem.

We're dependent on voting to determine "best" or "most important". Even if the sampling were statistically valid, populism does not equate to quality.

We should have automated systems for scoring articles on vectors of likely importance, relevance, accuracy and literacy so we can apply these to aggregators in conjunction with basic populism.

Because populism is easily gamed and has many known degenerate cases that are not fixed by throwing more populism at it.

Should HN offer to try to show us things that people with similar interests to our own found interesting? Or is that ranking style too prone to filter bubble issues and too similar to what other sites already do?

I would enjoy having a separate view that tries to rank content by what others with similar voting patterns have upvoted. Though it might not work very well with comments because I often upvote comments merely because I thought they were unfairly downvoted, not because I liked or agreed with them.

I've also tried to upvote comments that I thought made an interesting point that I disagreed with, which I suppose doesn't necessarily mean that I'm especially interested in comments by their authors.

I think you are presupposing that automated systems for articles can't be gamed . Google has to constantly update their search API to combat this .

Absolutely, it can all be gamed and any approach will have cases where it fails. That's why a balanced set of approaches using a handful of very different metrics is important – and possibly offering users the ability to browse by each metric separately.

What you're talking about is called an "editor", and web utopians spent the last decade or so screeching about how they're all autocratic gatekeepers that should be sent to the guillotine. It was a bad idea. We should have more editors.

No, editors editorialize and leak all of their human biases into determining what the readers see.

Do you rely on expertise in any domain? Quality editors are experienced in the domains they curate and edit. Given the volume of content that's produced, it makes sense to leverage editors. There's no bias-free news regardless, just as you point out, it's part of human nature. Finding a quality editor is well worth it, as you no longer need to spend as much effort in searching for and evaluating individual articles.

Such a system would likely help improve quality, it's just that it'd also be extremely difficult (and potentially costly/time consuming) to implement. After all, likely importance and accuracy are also (to some degree) community specific as well as relevance. A forum based on politics might consider a Donald Trump memo leak as the most important submissions whereas one about video game speedruns might consider a TAS video as theirs. A forum or site that supports a certain candidate might (unfortunately) want articles that support their preconceptions voted up more. And heck, even the way said factors are weighted would be context dependent. A news site striving for accuracy or a skeptics community might want to focus more on the submission's accuracy or literacy while a subreddit may want to focus on the likely importance of said submission.

It's an interesting idea, and I would love to see a system which could weight factors like this in a way that's easy to tailor for different communities. But when you consider the time and effort needed for such a system, I can definitely see why most just go with typical vote counts for ranking stuff.

It feels nice to spot something on the new list, give it its first upvote, and then the next day see it on the front page.

I've also seen a lot of interesting stuff on /newest that never makes it to the front page. Sometimes stuff that I personally find useful, even if it isn't relevant to the community at large.

Also, pro tip: page 2 has new posts with a handful of upvotes where one or two more votes can easily propel them to the front page. You can make a difference!

True -- at the right threshold one upvote can mean the difference between a post sinking without a trace or making it to the front page and gaining many more votes and views. In that way one upvote can trigger much more than 100 views.

"The importance of a single vote is easy to illustrate: On Hacker News, a popular post can receive 500 votes, which can generate ~50,000 visits depending on how interesting the title is. (I’m pulling data from Algolia.) Using simple math, this equates to"

Not even the half of it. I ran an experiment on /newest once, and just a single random upvote (mine) had a substantial effect on the odds that a submission would reach the front page, and since a front page appearance is worth anywhere from 5-20,000 pageviews, my single upvote was driving thousands of pageviews! And I've noticed when submitting my links and then browsing /newest for other meritorious links, this phenomenon doesn't appear to have gone away.

That's what I was talking about here:

https://news.ycombinator.com/item?id=13677501

Same thing seemed to happen to me whether coincidentally a bunch of simultaneous votes or some algorithm based on karma. I know I was more careful with upvotes on new stories afterward.

Yes, it's possible my high-karma makes my votes worth more. In retrospect, I probably should've done it with a lightly-used account. On the other hand, the implications for me are still the same - I know I take more time to do upvoting on /newest than I did before!

I just sent 100 more visitors to your website. (And it took way more than half a second - I had to read the entire post before doing so.)

Btw, tldr; "I know it's fun to dismiss social networks as frivolous, but link aggregators are a special type of social network that we should value higher."

I always felt "down-voting" should cost something. Maybe 1 karma/point/etc. so that people who would down-vote should feel they are invested enough to do so. As it is right now, down voting is too easy and detached to other's thoughts even after reading it.

Fundamentally, this is to create some value towards the point system so that people would put more thought into posting as well as down-voting. If someone wants the ability to down-vote, then they should write meaningful comments to gather the karma needed to do so.

I entirely disagree. Down-voting is a defense of the commons. The best way to teach us not to post shit, is for our shit to be down-voted mercilessly, with no explanation. I've got over 14k internet points, and I get down-voted every week at least, usually without discussion. I appreciate that, because it helps me learn to contribute better, without sucking me into a black hole of personality conflict. Every down-vote represents someone who was annoyed by what I posted, and we all benefit by reducing annoyance here. Truly unfair down-votes are typically corrected (...and then some) by up-votes from other users.

One probably doesn't understand why one is down-voted, the first time it happens. One probably doesn't understand every down-vote one gets, even with long experience. Over time, however, one gets a better and better idea about what we need here, and what we don't. It's even better if it's unconscious, and we edit our posts to be more "HN-style" without paying attention to why a particular expression conforms to that or not. This process is only disrupted by lengthy back-and-forth discussion of who said what and why and who interpreted that how and why.

Try not to think of down-votes as a reflection of your character or worth as a human. Rather, they are a snap judgment of a single verbal expression, made at a single point in time.

I'm not sure I totally buy the concept that 'good' up/down-voting saves people time. I rarely come to hacker news thinking "I'll just check the top n posts, filtering out posts below a certain score threshold"; typically, I'm thinking "What are the posts I can spend X time reading instead of doing something else?".

It's very ok to abstain if you have no informed opinion or interest.

The flaws of the system won't be fixed by gaming it even more.

That this type of information retrieval is portrayed as mankind's last hope, makes me miss Google Reader even more than already.

For lack of an better alternative HN is more of a baseline I check my twitter feed against.

Amusingly, more people than normal who've read the blog post are going to upvote this post.

Same for your comment. :)

Original comment was it was oldest one 2nd from the bottom. HN voting couldn't be that bad. Situation changed by time it refreshed lol. Maybe just cuz it's a recent topic.

I think I agree with the sentiment, but try to ignore the statistics.

I think the average user is more than 10x as likely to click a 1k link than 100pt.

Personally, if it's massively upvoted I'll click and skim even if it doesn't sound like it will interest me - if it's that important to the community, I 'should', right?

I agree, sometimes I think "how on hell this can have X invited?" then when I read it it's usually more interesting than the title and domain make it sound (and/or comments are where's it's interesting).

I feel that ranks should be a combination of explicit and implicit activity.

If a story link is followed a certain number of times, if a certain amount of time is spent on a page, etc. these indicators might contribute to an “implicit upvote” for a story or contribute in some way to its rank regardless of explicit up-votes. It’s also really important to have a good way to recognize and consolidate duplicates and reward them equally.

Activity in a comment subtree should also implicitly increase the rank of a comment. I have occasionally seen comments that were mysteriously not upvoted even once, despite spawning multiple levels of subtree discussion; in such case, isn’t the comment implicitly “more valuable” by creating a lengthy discussion, even if no one ever explicitly upvoted it?

No, a lack of upvote on a comment usually means the comment was wrong according to someone and it spawned an argument. I definitely don't want those cases ranked higher than good discussions where all commenters are receiving votes.

Similarly, an article that someone spends 10 minutes reading and then doesn't upvote (or comments without upvote) means it was probably a massive waste of time.

Not if the only discussion is to explain why it's wrong or off-topic.

Funny enough to shadowfacts comment, I think when you remind people that you can share/upvote/promote, you generally get a higher return rate. At least, that's what common wisdom seems to say. For example, if at the end of a blog post you say "tell us your story about x" where it relates to the post, you're more likely to get responses than if you didn't. This of course doesn't help bad content that you don't reach the end of.

No. Don't even try to guilt me into action.

The current landscape may favor you more if I take certain actions, but that does NOT obligate me at all.

Why haven't any link aggregator sites tried automating the voting system?

Take something like view time and commenting/opening the article as a vote and perhaps still have a flag function to help remove off topic or illegal content.

Manual voting seems like such a ham-fisted approach at finding what people want, and it disproportionately rewards those who vote over more passive observers.

Maybe I am out of date, but to the best of my knowledge, with my usage pattern I can't upvote in a meaningful way.

I am assuming it's still correct that upvoting on /newest is the most effective (or only) way to do it.

Just looked this up, I joined this site 2061 days ago, before that I was a logged out reader for a while. I don't think I visited /newest more than 10 times - I only read via RSS, and usually only once or twice per week - skimming over a lot, reading the comments for many - usually 100 - 300 posts at a time. If HN ever stops providing a RSS feed, I am probably gone anyway.

This is not a critique towards HN, but the OP maybe takes his own usage pattern as the default (maybe it is, I wouldn't know).

There's no evidence supporting the idea that every upvote brings additional 100 visitors to the upvoted link. Even if there's such evidence, it's not immediately clear from the article. The author doesn't consider the fact that some upvotes don't increase the point counter. There are other major flaws. The article doesn't disclose methodology used to obtain the results. It just makes the claim that an article with 500 points generates 50,000 visits, so every point is 100 visits, and expects us to believe it and take it as a fact. The author doesn't bother to prove that fact.

There also seems to be a mechanism where high-karma users boost a story more than one point. It would support Julian's point even more. I was at one point just boosting stuff in New that I thought people would be interested in across various topics based on prior front page results.

There also seems to be a mechanism where high-karma users boost a story more than one point

I've never heard of such a thing. What makes you think high karma influences the value of a vote? Have the mods talked about it?

Ive noticed a link in new that stays at 1pt a while go up a ton after I upvote it. This means either it stayed dead followed by a bunch of simultaneous upvotes or the site gives extra weight to high karma users. I dont upvote enough stories to be sure which it is. I just see points go up tens of votes second I upvote most submissions.

Most things in new don't get any votes at all so you're going to see lots of volatility there (nor does the ranking there matter much). The ranking computation is also not necessarily instant although maybe it is, I don't know. But I don't think this is any evidence of high karma users having some kind of supervote. If you could move items on the front page with a single vote (or with your mind), that would be another thing altogether.

Your last sentence happened a few times. It's what originally got me wondering. Let's also remember they have routines looking for people gaming the system. That could affect both timing and what effect a karma boost would have in visible results after a vote refreshes.

Upvoting on Facebook definitely does impact distribution. More likes translate to more people seeing it in their feed. This of course has a cumulative effect and leads to more re-shares. Virality is highly dependent on initial likes.

why not get 1 upvote automatically just upon opening/viewing? (please forgive me because i do not understand the entire process(and did not search to find if there is one))

An upvote is a recommendation, which should be based on reading the submission. The upvote should ideally come after opening/viewing if the submission is good enough. (Am I understanding your comment correctly?)

That would reward excellent (or clickbait) titles, which might not be the good articles.

For members, change HN to display new posts as the default page, replacing the "new" navigation item with "top"

indifference is voting.

Link agregators have a free rider problem. Just like democracies without mandatory voting.

Those two are not even remotely connected and to claim a lack mandatory voting implies a free rider problem is ridiculous.

Reply on news.ycombinator.com