Design: Voting up or Down is Dead

http://www.whattofix.com/blog/archives/2007/10/voting_up_or_do.php

by DanielBMarkham • 19 years ago

40 77 19 years ago

aston8• 19 years ago

I voted up because I agree. Would I have done so if I disagreed, despite it being decently well-presented? Should I have?

So here's the question: How do you do better?

DanielBMarkham9• 19 years ago

I would form a startup that had a better idea and I would try to adapt it to make more users happy!

Not trying to be a pain, just that it's a big question, and whoever makes progress in this area is going to be accomplishing something big, in my opinion. I have my ideas, which I'd like to try out in the market and not kick around on a board. Hope that's an okay answer. I'm not a "cool idea makes the business work" guy, but neither am I a "give away the store" guy either. I think there's going to be some proprietary magic going on when this nut gets cracked, much the same as when search was cracked there was a lot of behind-the-scenes magic that Google did.

BTW -- thanks for the vote. Karma whores unite! :)

pg6• 19 years ago

I think you must have arrived since the explanation of how we're going to avoid drift.

http://ycombinator.com/hackernews.html

Vote weighting isn't turned on yet, but for the last couple weeks I've been thinking it was time to.

Incidentally, the poll problem can easily be fixed by changing the frontpage ranking algorithm to treat poll votes as worth e.g. a third as much.

DanielBMarkham9• 19 years ago

Awesome!

Color me skeptical about stopping drift, though. It'll be neat to be proven wrong. I wonder how you'd prevent the system from just emulating the editors, instead of the audience. Now with YC perhaps that's exactly how you'd like it, but for a more general purpose site I'd want more of a user-focus than an external-editor-focus.

For the newer people, thank you for the link about what the appropriate topics are. I think a time or two I asked "what does this have to do with hacking?" Now that I know the larger focus, I'll be more reticent.

sethjohn2• 19 years ago

It does seem like a "weighted" voting system defeats the purpose of having a voting system at all. What's the point of asking everyone to give their opinions if the only opinions that will matter are those that conform to the opinions of the editors!?

Of course, in truth I might be quite happy with this system. I've liked what the editors have let through so far (though last few days less so) and I would probably rather read what they suggest than what the "group" suggests.

Alex3917• 19 years ago

Before you change the ranking, are there any specific things you like or dislike thus far? I've noticed a lot of new user accounts lately, so maybe now would be good for a state of the union.

pg6• 19 years ago

Stuff that takes less than 30 seconds to view and vote up is a bad sign. In fact, that suggests a solution...

earthboundkid• 19 years ago

I think one thing that might help is if the choice was a little less binary. A one to five star rating system might help sort out the difference between "a decent way to kill some time at work, so up vote" (3 stars) and "something that will actually change the way that I look at the world, so up vote" (5 stars). Something to think about.

ijoshua3• 19 years ago

"Maybe" is one option too many

http://www.zeldman.com/2007/06/20/remove-maybe-from-invitati...

dcurtis2• 19 years ago

A better solution than that would probably be to add two different types of votes, "agree" and "disagree." Maybe even "interesting." That way you don't have upvotes meaning a bunch of different things.

Then you can indicate the level of agreement/disagreement on some sort of graphical/iconic bar. Having a number of points means almost nothing except that a bunch of people clicked the arrow. With an agree/disagree/interesting system, you can even do math in the background to determine the kinds of articles that certain people tend to like, and you can really figure out their personality that way.

Then again, polarizing things into three categories isn't a very accurate way of discovering what is truly popular in a community. But I do think it's better than selecting certain people to have heavier votes.

almost• 19 years ago

Please do! news.yc seems to be veering a little off track a bit lately and it'd be nice to see it come back.

After all, if people want the full wisdom of the crouds/mob rule thing then there's always Reddit.

sbraford2• 19 years ago

Not to nit, but communities like reddit aren't really following "wisdom of crowds" model anymore, as outlined in the book.

To truly obtain wisdom from the crowd, it has to be collected:

a) non-sequentially (i.e. all votes are cast independently of what everyone else cast)

b) taken from wide cross-section of the population.

Once a community has formed around particular topics (far left politics, atheism, xkcd, "vote up if ...") you may be collecting wisdom of _that_ crowd but not wisdom applicable to anyone beyond that select group.

fauigerzigerk• 19 years ago

I have severe doubts about this kind of voting system, weighted by moderators or not. A post that was voted up by 100 people and down by 100 people (assuming equal weights) will look exactly the same as one that nobody ever touched. It ruins the effectiveness of voting as a ranking mechanism.

DanielBMarkham9• 19 years ago

DZone, which is linked in the article, has up and down votes that they keep track of separately. That way, they don't cancel out. Instead, you get an idea of the total positive and negative sentiment.

ijoshua3• 19 years ago

PG, I'd be interested to know why you chose to implement a voting/rating system for stories on news.yc. That is, did you consider a "weightless" submission system like, for example, MetaFilter, where front-page postings are a privilege that every member has equally. I've been a member of MeFi for several years and the effect of "gaming" has never really been an issue there, because there is no reification of trustworthiness in the form of a "karma" score. Certain members there certainly have name-recognition based on their history, but this gains them nothing in terms of influence on the focus or direction of the site as a whole.

dyu• 19 years ago

I think up/downs are to judge the worthiness of the link, and agreements should be expressed in the comments.

DanielBMarkham9• 19 years ago

Would you vote up for a link to a story that you violently disagreed with? Be honest now. And if you would (I _think_ I would) do you really think that's the way others are voting?

reidman3• 19 years ago

I just experienced this dilemma with the 'Facebook Sucks' post by Dave Winer (heh) that went up on YC news earlier today or last night.

I thought the linked story was pretty lame, but at the same time, I thought it was an interesting conversation. In the end, I didn't vote up the link, but I did participate pretty heavily in the comments.

I would like a way to distinguish between liking the story and thinking that other people should see it. But at the same time, I realize how complicated that would make things.

This question/dilemma actually relates directly to my YC app/idea, so I'm going to be following the comments in this thread pretty closely.

kmt• 19 years ago

I would. By voting up a link or a comment you improve their visibility. This should bring more people to the discussion and that's good if you are passionate about the topic (no matter if you agree or not, or not quite sure).

utnick3• 19 years ago

the problem is that even though WE don't like poll stories or random bush bashing on the front page, a TON of people do like them obviously, and you need to accomadate them too

anyone ever used pandora.com?

you give it an artist, it plays you a song by the artist or in the same genre, you vote the song up or down, eventually pandora learns enough about your tastes to only play songs that you enjoy

what if a social news site kept track of what kind of stories you voted up or down, did some magic to draw connections between types of stories, so the front page of the site would look different for each user... people that don't like poll stories wouldn't have poll stories on the front page, etc, etc

pg6• 19 years ago

even though WE don't like poll stories or random bush bashing on the front page, a TON of people do like them

Not quite, unfortunately. The problem with voting, as the reddits realized from the very beginning, is that people will vote something up not just because they liked it, but because they want other people to read it. And a site full of what the majority want everyone else to read could be significantly less interesting than one full of what people like themselves.

And with polls there's not even a pretense that they reflect anyone's interest. The voting mechanism is being hacked for another use.

aston8• 19 years ago

People voting for whatever they want other people to read is basically the same as voting for what they want themselves. I know surveyors occasionally ask questions like "How do you think most people would answer the following question: ___" in order to get an honest answer in cases where people might not want to fess up to some position personally.

greendestiny• 19 years ago

Individual recommendations don't fine tune social news, they destroy it. You no longer have a shared resource and the communication becomes very indirect.

utnick3• 19 years ago

If you make the magic good enough, the filtering system will only filter out stories that the reader wouldn't read or comment on anyways. So it won't really affect the community.

samson• 19 years ago

I agree, systems like digg and alot of other voting sites, don't seem to create alot of invested time for the users.

What I mean is, the data they have the opportunity to collect and use to benefit the individual would make users infinitly more loyal, and reduce the chance of them being overly liberal in their voting habits. The reason being is that they'd CARE. My argument then if I had one to begin with is that users don't care about the system because there's nothing invested in it for them. Yes..Yes..the community..I know, but whats in it for them to make them care.

waleedka9• 19 years ago

It's not a voting problem, it's a scaling problem; I believe. Simply put: social news sites don't scale very well.

As communities grow:

1. You and I can't get things on the home page any more. It becomes only for the well connected.

2. Subjects lose focus.

3. It becomes a gold mine for spammers. They'd do anything to game the system if they can get 234,000 visitors from one post [1].

The solution, I believe: Keep communities smaller and focused. If need be, create many small communities.

[1] http://blog.mindvalleylabs.com/marketing/how-to-get-traffic-...

aston8• 19 years ago

At the same time, monetizing eyeballs requires lots of numbers. I guess the solution then is to make a site that feels like a ton of people are there, but still operates quality-wise as if there's a small community actually controlling the content.

waleedka9• 19 years ago

True; monetization requires big numbers. But why not split them into smaller communities? Rather than one huge home page, have a lot of smaller ones. And, then, allow the user to select the 5 or so that they like to participate in.

Restrictions are sometimes good. Compare Facebook to MySpace. With MySpace it's one huge community; everyone can see everyone. With Facebook, you can only join a small set of networks. You cannot even join San Francisco and Silicon Valley at the same time. True, people hate it, but it keeps the system manageable as it scales.

This is a subject I'm interested in and trying to build a startup to address it.

bct• 19 years ago

If you decentralize you don't need to monetize. (It's good for everyone except entrepreneurs).

I've got plans in the back of my brain for combining linkblogs, the Atom Publishing Protocol and something like http://www.intertwingly.net/blog/2006/09/03/MeMeme into a decentralized version of reddit/digg/news.yc, but I already have too many projects on the go.

The really nice thing about the idea is that it allows for a lot more experimentation with the best ways of aggregating and sorting through the massive quantities of URLs that are available. Right now only big players like Reddit and Digg and have big enough data sets to really experiment, and they're naturally reluctant to mess with a formula that's working for them.

reidman3• 19 years ago

Since my YC app deals with this directly, I've been thinking about it a lot. My idea has to do with voting quotes up/down (kinda like Jyte, http://jyte.com ).

In allowing people to vote up/down, I realized that there needed to be a separation between 'Yes this quote is accurate' and 'Yes I think this quote is important', since the two can be exclusive.

In the end I decided to implement a two-step process. The first step is to Agree/Disagree with the quote at hand. Once a user has voted in this way, they have an optional second step to vote for how 'Important' the quote is (important or unimportant).

I have yet to see how it will work out, but I'd love any comments you guys could provide.

aston8• 19 years ago

I'm not sure if I agree with the idea that importance is completely independent of agreement. It seems to me that agreement is one of the many variables that play into what's important (along with timeliness, 'interestingness,' widespread appeal, and tons of other things).

I think the quality not scaling is the main issue. Adding another step to the process is just going to make people less interested in contributing their opinion at all.

adsyoung• 19 years ago

I'm also very interested to see how something like that would play out. I'm kinda skeptical of any solution that adds complexity to what the user sees though. I'm guessing the more you ask on their part, less people will make proper use of it and click on all the required things appropriately. Then again perhaps thats what you want, only the serious users doing the clicking... We need experiments...

waleedka9• 19 years ago

By doing so, you make the UI complex. Only a small percentage of people vote in social news sites as it is, even with the simple one-click up/down.

reidman3• 19 years ago

I've designed it so that the user doesn't see the voting option for importance until after they've voted for the accuracy of the quote.

whacked_new2• 19 years ago

My previous YC app dealt with this and I posted various comments already, but I still like this topic.

An introductory study into conversation theory will tell you how theoretically bad up/down is. It seems to me that the two-step process is still broken though, because "important" is not a well-defined word. The best system I have seen online, which actually does a decent job of utterance categorization, is Slashdot. Based on the Gricean Maxims and you can come up with some core cirteria for any given utterance, and /. actually covers them surprisingly well. Obviously it's still not a perfect implementation, but they have done a good balance between categorization, ease of categorization, and evaluation of categorization. They really need to grow out of the +5 limit though.

While /.'s criterion are good, they don't cover enough ground to become a universal system; it doesn't scale to other cultures. There's a lot more to this I think, but I'm just be rambling again.

yters4• 19 years ago

In my opinion, this is as strong AI problem, and only people have strong AI. Therefore, large social sites are always going to need human moderators to have a good signal to noise ratio.

aston8• 19 years ago

That's a bit generous, I think. Most people aren't smart at all. And even the ones that are don't act objectively and without bias. Ever.

Although I will admit that moderated solutions almost always come out better. See slashdot's complete lack of this problem.

yters4• 19 years ago

I didn't say it's a complete solution, otherwise we'd never have disagreements, wars, etc. But, it's the closest we'll get to a solution.

That being said, I think algorithms can be very effective as intelligence amplifiers, allowing us to make higher level decisions and micro manage less.

Xichekolas• 19 years ago

Slashdot may not have the problem on a story/post level, but have you ever tried to read every comment on a story? A third are hate speech or spam, another third are repetitious or illogical, a sixth are insightful/funny, and the other sixth would be if they didn't happen to be against the wider political views of the people that had mod points that day.

I think the only real solution is to have a small community whose members are actually heavily invested in the topic... basically, experts.

If I try to go add to a discussion about some new Myspace feature, my thoughts are likely going to be off topic, misguided, or an outright waste of space... all because I don't have a Myspace account, and never have. For the same reason, if a bunch of Joe Sixpacks show up on Hacker News to discuss startups, they are probably going to bury any thoughts from actual startup founders or people involved in the 'scene' (whatever the scene is).

I'm not sure how you maintain that, short of making a community invite only. Even Hacker News is starting to show the pain of size. The front page often has pretty off topic stuff on it, when I, for one, came here originally to keep up on the startups funded by YC and the thoughts of the people involved in them.

Sure, PG still comments here regularly, but how long until this place grows and he becomes an absentee owner like CmdrTaco? He may not stop commenting, but there will be so much crap in the comments that few may notice what he says.

ijoshua3• 19 years ago

"Most people aren't smart at all."

Very correct. In fact, nearly half of all the people you'll meet have a "below average" intelligence.

alex_c• 19 years ago

Seems to me like Slashdot's comment system does two things that might address many of the problems you mention:

- instead of "Up" or "Down", the voting is a bit more granular - "+1 Informative", "-1 Flamebait" etc. The UI isn't MUCH more complicated - instead of a "vote up / vote down" link you have a dropdown.

- Voting points are a limited resource - assigned to "reliable users" randomly and expire after a few days. This makes it a bit harder to get a core of power users who are practically guaranteed to be voted up. As a bonus, the scarcity probably means that people are more likely to use their votes "wisely".

I wonder if something like this has been tried for selecting the front-page stories on a social news site, and how well it would work. As far as I know everyone's following Digg's model.

DanielBMarkham9• 19 years ago

I'd like to see something like this tried instead of the yes/no button -- pick something with a little more meaning. Then at least the system has more information to present.

whacked_new2• 19 years ago

Yay for mentioning Slashdot!

Slashdot has a good system, but for increased granuarity users actually need to think more about what they're voting, instead of what is possibly a subconscious, wholistic, instantaneous judgment, which seems pretty applicable for Digg. So there's a tradeoff for a system to be "easy to use" rather than "intelligent."

In this sense the UI becomes a bigger challenge than the data processing. See http://x-o-o.com/news.php?sort=new It's a dictionary-supported rating system with granular voting, in the sense that you can vote 10 or 0.1 if you choose to do so; it also supports weighted voting.

It is much harder to use than digg though so such a system probably will not reach popularity. But I think theory wise it is the best implementation.

toisanji• 19 years ago

Slashdot's commenting system is one of the best I've seen. One thing I wish they had though is the ability to filter out types of comments. You can already label comment as "funny","interesting", or whatever, but you can't filter by those labels. I want to be able to read at a threshold of 3+ and not funny.

mdemare• 19 years ago

What I liked about Slashdot was that you could redefine +1 Funny as -1 Funny.

JulianMorrison• 19 years ago

I see Slashdot in a darker light. To me it illustrates two principles. First, that there is a level of "firehose" beyond which quality-maximizing editors don't scale. Second, that an abstract judgment (up/down) is less abused than a list of judgments, all of which are necessarily trite. Handing the right to judge out by lotto only broadens the standard deviation of voters - if anything, it worsens the groupthink. "Power users" have more invested in a vote.

Thinking about what makes a vote worthwhile, there has to be some high barrier to entry (or any troll will pile in) and it must not itself be subject to groupthink (or it will result in an automated clique). In the case of Hacker News, I propose the metric: anyone who has founded or co-founded their own startup (defined as: in at the start, and has partial ownership) may vote.

angryearthling• 19 years ago

i don't think it is a lotto as to who gets to vote. those who participate in moderation and those who make comments that are voted up get to vote.

i think one of the very good things in the slashdot model is the cap at 5 on votes. it lowers the number of people who go out of their way to karma whore. what's the point when all you will get is 5 points?

i've used slashdot for a long time (my user id is 5 digits). every year new sites come along.

i used digg for a year before it degenerated into the mess it is now. it went downhill fast once it expanded from a tech only site.

i used reddit for 6 months. it has gone the exact same way. as i type this it has been offline for a day.

after giving up on these i always drift back to slashdot. maybe the question is why do i leave slashdot in the first place? what is wrong with slashdot that i look for other sites?

apgwoz• 19 years ago

The greatness of up/down is that recommendations are simpler, though arguably reddit's recommendation system doesn't work... so, I'm on board .Slashdot moderation!!

udfalkso• 19 years ago

Explicit rating & ranking will inevitably crumble under its own weight. You have to use implicit data that isn't easily subject to collusion. Look for trends among disjointed groups that are acting in their own self interest.

I wrote about this a while back: http://breasy.com/blog/2007/07/01/implicit-kicks-explicits-a...

kn0thing• 19 years ago

To the argument about reddit suffering from "Vote up if..." syndrome, we've thought about building a "poll" submission type (like a self-post) that would create an actual poll on reddit as the submission. We'd like to see fewer of these polls on the front page, too, but still give worthwhile poll a chance -- they're just in the minority.

We figured this would make the poll have to earn the "up vote" instead of just sucking people into clicking up because they "want Bush in prison." It could even graph the results! Histograms are sexy.

Incidentally, newsvine has something like this. We're in the middle of getting new reddit online, so who knows when/if this would actually happen. Perhaps voting up/down will already be dead by then ;)

Readmore• 19 years ago

I had up and down ratings on http://www.klipboardz.com and then decided to switch them out for a system that just rates on number of times the story is read and the number of comments. I think it's a better system than most social sites. The only problem I expect is burying content, but that isn't a problem since the site is so small.

sbraford2• 19 years ago

This is why for a new project I just launched I'm using pageviews as the metric (it's a pic sharing site) instead of voting up/down.

Each IP address can only register one unique "vote" per 24 hour period. This makes the system still vulnerable to proxy servers and botnets, but it's not nearly to the point anyone would want to game it yet. (blacklists will go a long way against proxy server attacks)

Design: Voting up or Down is Dead

Related Stories