The social media revolution is here, and it is here to stay - everybody and their dog say so, after all! But compared to the endlessly swelling user numbers, technological progress of the various social media platforms is lagging by whole astronomical units! People are still enthusiastically discussing how swell it is to be able to up- or downvote content (or users - we call it "karma"), or to be able to tag stuff!
I say the current state of this discussion is disheartening - nay, positively depressing!
So you ask me: Why, what is your beef with all those great tagging and voting mechanisms?
In two words: They suck!
Two more words: Big time!
One question: Why?
Just consider the following situation:
[Danger: Metaphor ahead! We know they all begin to leak at some point, so don't press it!]
Imagine you want to sell steaks. You are a bright guy, so you invest in some market research about your potential customer base. When the results come in, it turns out that half your potential customers love their steak with a big glob of ketchup on top and half do not like any ketchup at all.
In this situation, would you seriously suggest to serve every steak with a small glob of ketchup? Of course you wouldn't! Why? Because you can plainly see this is not a monomodal but a bimodal distribution, and that by just averaging everyone together you will appeal to neither of the two groups!
But wait, isn't this exactly what voting mechanisms currently do?
You bet it is! Count the votes, average them out, and then base content recommendations upon them!
What a bloody shame and disgrace!
People, we have computers! You know what they are good at? Crunching numbers!
So, for example, do not just count how often someone tagged something as "funny" or upvoted something, but save that info including the user's ID. Then, if user A looks for funny links, let tags from user B (who's "funny" tags are highly correlated with A's) carry more weight than tags from user C (who prefers a different kind of humor). What you will get is a tagging/voting mechanism that is lightyears ahead of what everybody else is doing! The reason is because it won't drag everything down to averages, but be able to identify and sustain multiple parallel subcultures in the user base, which have different preferences!
There are a myriad ways to achieve something like this, and get over the dictatorship of mediocrity!
All you clever, tech-savvy social media entrepreneurs: Have you not ever heard of multimodal distributions?
All you multi-million dollar heavy social media companies: Have none of your (probably) highly paid experts?
So you want to be the next Facebook, the next Twitter, the next Big Thing?
Look no further! Do something worthwhile! Read up on statistics! Know some statistical machine learning! Know your clustering algorithms! And be it freaking k-means! Go out there, and create my next favourite social website!
Do it now, or you'll be late!
Btw I've already ranted about this on hacker news and in a reader comment on sixrevisions (awaiting moderation). Parts of these rants have been re-used here.
