On a site like Reddit, Hacker News or Digg, a link gets ranked based on how well the average user likes it. A system like this works well when the users all have a similar idea of what is interesting. However, it's difficult for a site to keep this property as it grows, and eventually the average user ceases to represent anyone in particular. This is the dilution problem.

Thought Ocean uses a different system for ranking posts, a system that gives different results for everyone, and works even when you aren't represented by the average user of the site. It is based on how current flows through circuits.

You need to give Thought Ocean an extra bit of information, and it uses this extra information to rank links more intelligently.

The information you give Though Ocean is:
a. The posts that you like
b. Other users of Thought Ocean whose judgement you particularly trust

How does Thought Ocean use this information?

Suppose you know someone who uses Thought Ocean. I'll call them User A. You tell Thought Ocean that you particularly trust the judgement of User A. I'll abbreviate this by saying you vouch for User A. This is represented in the diagram by an arrow. User A likes a certain post, and indicates this to Thought Ocean. I'll also call this vouching, and it's also represented by an arrow. So you vouch for User A, and User A vouches for a post. This is some evidence that the post is good.

Now suppose we put another person in between you and the post. So you vouch for User A, who vouches for User B, who vouches for the post. This is still evidence that the post is good, but it's weaker than in the previous situation. You don't trust the judgement of User B quite as much as that of User A, because you don't directly know them.

The more general law is:
If there is a chain of people from you to the post, that is evidence that the post is good. However, the longer the chain is, the weaker this evidence is. More in series = worse.

Instead of adding User B in a chain with User A, what if we add User B beside User A? So you vouch for both User B and User A, and they each vouch for the post. Clearly, having an extra person vouch for the post is stronger evidence that the post is good. And if we were to add more people like this, that would be stronger evidence still.

So now we have two laws:
a. More in series = worse
b. More in parallel = better

If you know anything about circuits, this might seem familiar to you. Circuits obey similar laws.

With constant voltage,
a. More resistors in a series circuit = less current
b. More resistors in a parallel circuit = more current

Because of this similarity, a ranking system based on how current flows through circuits will have the properties we want it to have.

Imagine all the users and posts on Thought Ocean are nodes in a circuit. Wherever one user vouches for another user or post, there is a one-way resistor connecting them. It has to be a one-way resistor because vouching for someone shouldn't imply that they vouch back for you. Now attach the node corresponding to you to one terminal of a battery, and take all the nodes corresponding to posts and attach them to the other terminal of the battery. Current will start to flow through this network of resistors. Thought Ocean simulates this process and ranks posts according to how much electric current flows out of them.

To rank posts for a different user, it would be their node rather than yours that gets put to the battery, and as a result current would flow through the network differently. So Thought Ocean ranks posts differently for every user.

A user of Thought Ocean, by default, does not have any power to affect how posts are ranked for you. If they do have this power, it's because someone granted it to them by vouching for them. So it's possible for the website to be used mostly by people who don't share your taste, and yet still have your rankings determined mostly by people who do. As a result, Thought Ocean can solve the dilution problem.