How to break zipf's law?

Does anyone have any information or some kind of idea which could ignore or break zipf's law in a social media platform, I mean Sup Forums completely breaks it by no system of voting but is there a way to enable voting and still bypass the law?

What are you asking?

>Does anyone have any information or some kind of idea which could ignore or break zipf's law in a social media platform

what does that even mean? Zip's law is not functional in a social media platform. Most people would not even know what that is and it would likely be used wrong.

I guess you could make memes that creates a new meaning to zipf's law so that it can not be used in a meaningful way and therefore break.

>I mean Sup Forums completely breaks it by no system of voting

No it does not.

>but is there a way to enable voting and still bypass the law?

Use whatever poll system that Sup Forums uses.

There is a system of voting in 4 chan. Popular threads get bumped, unpopular threads get 404'D.

No, I mean look at reddit, 20% of the users get 80% of the karma, 20% of the facebook pages get 80% of the likes, same with twitter in terms of followers, instagram? what not? So is there a robust way to engage some kind of voting/liking/following that would help distributing the general interaction evenly?

>I mean Sup Forums completely breaks it by no system of voting but is there a way to enable voting and still bypass the law?

You can rank posts by the number of replies.

Atleast comment section is still far from being followed by it because of the lack of the voting system? But the question is, that is there a nice robust way to bypass the law with the use of any kind of voting?

Conjecture: If you plot the number of posts a thread gets before it dies for all threads, the graph will follow zipf's law

Exactly! Everything which involves any kind of voting follows this principle, so can we do something about it?

>is there a nice robust way to bypass the law with the use of any kind of voting?
I assume that you would break zipf's law by making upvoted content harder to find, rather than easier to find.

Looking broadly at it, Zipf's law seems to emerge from self-reinforcing trend. The more popular something is, the more likely you are to stumble upon it, therefore it will get even more popular.

So if you break this feedback loop by artificially making it harder to find more popular content, you could probably counteract it to some degree.

I mean, probably? The law's been tested on way more complex things than Sup Forums. It'd be trivial to do the analysis.

>So is there a robust way to engage some kind of voting/liking/following that would help distributing the general interaction evenly?

Just use violence, force people to like/vote/follow unpopular shit by using violence or by killing people who vote for popular shit. it will make people think twice.

The more important question is whether or not this makes any sense. It would probably make sense for stuff where “upvotes” directly correlate to some other form of benefit, like when you get paid per youtube view.

Based on this, what might work is if you keep the popularity of content invisible - i.e. allow “up votes” but don't display the current score publicly and don't allow sorting/filtering by score.

In that sort of system, you'll still find zipf's law because of e.g. people sharing the URL to a popular post via third party communicaton channels. If you want to go further an counteract that, you'd need to make content difficult to link to - for example by making it so that no content has a permanent or static URL, but instead every time you view the page you get presented with random content.

So, basically archiving would be my best bet.

>Exactly! Everything which involves any kind of voting follows this principle, so can we do something about it?
It applies to anything that can be ranked not just voting systems.

An example of how this could work with a real-life example:

Imagine a music distrbution platform where artists can upload original music content, and people (paying subscribers) can randomly stream music; like a radio station - with no precise control over what they want to listen to (except perhaps the genre).

The only control you have is to press a button to skip the current song and move on to the next one. Every time somebody listened to a song to the end, it counts as an upvote. Every time somebody skips the song, it counts as a downvote.

At the end of every month, all of the incoming subscriber money is distributed among the artists in proportion to their song's score.

Since you can't choose what the next song will be, you can't “link” a song that you liked to somebody else, and therefore there's no viral / sharing effect. Since you can't see what scores others have assigned to the songs, you're not being biased by herd mentality / previous votes.

This system will probably break Zipf's law, while also providing a useful service (artists who make more popular songs - that is, songs which are truly more popular and not just gamed to be more popular via chart manipulation and aggressive marketing - get paid the most)

Wouldn't that be extremely detrimental for the website itself? I mean the more shares it gets the more popular the website would become?

The thing is counteracting and not displaying upvotes , any kind of stopping the content to flow analogy is just straight up disaster for the community and website's growth point of view, and hence I'm here.

Is there any crazy other idea which wouldn't restrict the flow of content and hence popularity of WEBSITE/COMMUNITY and still stop individual to get a lot of power?

>unpopular shit is now "popular" by force.
Great one

In fact, another thing you could add is an explicit “upvote” button for a song you really liked, to grand them bonus score.

>Wouldn't that be extremely detrimental for the website itself?
Yes, which is why nobody tries to avoid Zipf's law in practice. Instead, they all try to game it and make as much money off of shitty “viral” content.

That solves OPs problem in a easy and clean way. Well... Clean unless these fucks decide to be a bunch of smartasses and like popular shit, i'll cut them up.

Now this is fucking amazing! This is what I'm talking about, a crazy out of blue idea! Thanks I will reconsider and brainstorm around it,

I want to give internet something it deserves, and it definitely doesn't deserves this shitty cringe cancer.

Keep in mind your music sharing platform (or whatever) will be unpopular as fuck because people want to follow Zipf's law. I mean, even within my local music library, the number of listens for a track follow's zipf's law, and I have absolutely fuck all to gain from listening to the same track multiple time.

Obviously I ain't going exactly what referred but will try to find a loophole which could ensure the popularity and content have atleast a linear growth to something awesome.

>because people want to follow Zipf's law.

Which is why the "violence" idea is the only reasonable idea in this thread

Is anyone interested to do a collaboration? I mean I am developing a social network platform, it's almost done with typical content sharing and managing bullshit, but now I'm going deep into networks, as i am as called 'newfag' in Sup Forums so idk if revealing anonymity is just like a taboo in here.

>This system will probably break Zipf's law
No it wont. Zipf distributions are not only due to network effects.

This thread feels like one gigantic samefag.
Also the question is posed like shit.

OP, you should implement the kys voting system

This

Since you're new to Sup Forums, you should lurk a but more and figure out more about the place before asking these types of questions, IMO.

Especially this one
>Is anyone interested to do a collaboration?
is a pretty big taboo on Sup Forums. Not because it's frowned upon, but simply because it doesn't work. Also nobody wants to help you implement your shitty idea, and it won't work nearly as well as you think it will, so you're on your own either way.

Have fun wasting your time and money, though. Startups are a joke.

Would you mind elaborating?

You find Zipf law in the behaviour of all kinds of systems, both social and non-social. You can even find Zipf-like distributions in the organization of a pile of sand.
While the mechanism behind this is not understood, it is clear that network effects are not the only mechanism (I doubt that it is even an important part). The preferential attachment in power law distributions are properly a deep statistical bend to our reality.

Yes but I mean can you give an example of where you'd expect to find a Zipf distribution in my proposed system?

>is anyone interested to do a collaboration?
Sure lad, I'll start on the logo

In the frequency distribution of the songs.

Okay, I'll give you that one, but I don't think that's really what this thread is about.

HOLY SHIT
plotted the first 15 Sup Forums threads (for some reason i couldn't access more than that from the catalog's json)

The correct way to break zipf's law is to prevent the leader from doing anything more. If the lakers have more subscribers, then take some of them and give them to the bobcats

There's one way I can think of but I'm not sure if people would even like it.

Let's say we want to take Sup Forums's replies from this (btw there are only 15 posts because the catalog is for some reason divided into 10 sets) into something more uniform.

It can be done implementing posting currency and dividing the posts into low, mid and high popularity.
Posting in a high popularity thread has a cost of 3, a mid popularity a cost of 1, and a low popularity has a payoff of 2.

The problem is that this would lead into shitposting, but at least all threads would have a more or less uniform number of replies.

>I mean Sup Forums completely breaks it
Threads obey it by posts per id.

and by replies per thread

i wonder if images per thread fits the law too

I think this graph is distorted because of the bump limit, i.e. the most popular threads get less replies than they otherwise would have.

And that might indicate a good way of cubing zipf's law: making all posts/threads have bump/reply limits, so past a certain point everything just self-censors.

I imagine images per thread is going to be roughly proportional to replies per thread.

What would be more interesting to graph is the ratio of image replies to non-image replies. Does the distribution of those ratios follow zipf's law?

Yes, the first thread should have 400 replies from the law, but it only has 250 or so. But even capping the most popular threads you still have a lot of unpopular ones.

Right, but the matter of fact is that zipf's law is a 1/x distribution. In other words, as your number of threads increases, the popularity of the biggest thread increases proportionally.

So suppose you have 100 threads and the first thread has 250 posts. Now say your site grows to 300 threads in total; the biggest thread would grow to 750 posts in size. And so on. As your website scales, the popularity of the “top” content scales proportionally.

By introducing a hard cap X on the thread size, you can let your website scale to millions of users without letting threads catch up, so the resulting distribution would simply be thousands of threads with X posts each, and then a handful that didn't hit the post limit.

Of course, in practice, the effect we see on sites like Sup Forums is that people will simply “continue on” discussion in the next thread. Stuff like essentially never dies since somebody just keeps on making a new thread, so the real thread would be much, much larger. For a proper analysis, you'd have to piece together these “continuation” threads.

And ultimately, that's sort of why Sup Forums always had a bump limit - to keep threads small enough, since big threads are hard to browse, follow, catch up on, store and display.