3 ms·

I was thinking about doing something similar. One of the things I wanted to incorporate was an reputation ability (beyond just blocking). It would be similar to

I was thinking about doing something similar. One of the things I wanted to incorporate was an reputation ability (beyond just blocking). It would be similar to a referral engine in Netflix or Amazon. So, I'm user A. I mark user B as a high-value poster. User B marks user C as a high-value poster. When I run across a post by user C, I see they have a good reputation. If many people I ranked as high-value, also ranked user C as high-value, then I see user C as having a very high reputation. The same, in reverse, for low-value posters.

meemoo 13y ago
Are there any link voting sites on the public web that have tried this? I'd love it if YouTube comments worked like this. Why don't they?
- tprice7 13y ago
  I made a site a while ago that acts this way. Basically when a person vouches for another person or post it creates a metaphorical one-way electrical resistor from the person to the other person or post. Then it ranks posts by connecting you to one end of a battery and all posts to the other end of the battery and ranking posts by how much current flows through them. It is at thoughtocean.com. Pretty dead now though.
rolleiflex 13y ago
This is cool, but you're making assumptions about information that is outside of your immediate connections. This could work in a network where one could trust that all the information one receives from immediate connections about others is true, but that is not the case. In your case, user B could mark everybody as a high value poster, and it be successful in diverting your attention. This causes problems even in the core stack, I cannot believe what a node B will say about at what IP address node C is. A malicious node B could overwrite everybody's node information to one single IP address to create a massive DDoS. So the solution I found is that there are separate degrees of trust for derived and produced information. If you receive something over the network about a third party, that's derived information, which might not essentially be true. But if you have connected to that node personally, that's produced information, and no derived information can overwrite that.
- scottlilly 13y ago
  You're right, user B could make incorrect (to me) ratings about other users. Similar to what you say about derived and produced information, I would have "personal" and "external" ratings. My personal rating for user C (that I make after reading some of their posts) would override all my other external contacts ratings of user C. Over time, if I have a high percentage of different personal ratings from user B's ratings, then the system would apply less weight to his ratings for other users.

meemoo 13y ago

Are there any link voting sites on the public web that have tried this? I'd love it if YouTube comments worked like this. Why don't they?

tprice7 13y ago
I made a site a while ago that acts this way. Basically when a person vouches for another person or post it creates a metaphorical one-way electrical resistor from the person to the other person or post. Then it ranks posts by connecting you to one end of a battery and all posts to the other end of the battery and ranking posts by how much current flows through them. It is at thoughtocean.com. Pretty dead now though.

rolleiflex 13y ago

This is cool, but you're making assumptions about information that is outside of your immediate connections. This could work in a network where one could trust that all the information one receives from immediate connections about others is true, but that is not the case. In your case, user B could mark everybody as a high value poster, and it be successful in diverting your attention. This causes problems even in the core stack, I cannot believe what a node B will say about at what IP address node C is. A malicious node B could overwrite everybody's node information to one single IP address to create a massive DDoS. So the solution I found is that there are separate degrees of trust for derived and produced information. If you receive something over the network about a third party, that's derived information, which might not essentially be true. But if you have connected to that node personally, that's produced information, and no derived information can overwrite that.

scottlilly 13y ago
You're right, user B could make incorrect (to me) ratings about other users. Similar to what you say about derived and produced information, I would have "personal" and "external" ratings. My personal rating for user C (that I make after reading some of their posts) would override all my other external contacts ratings of user C. Over time, if I have a high percentage of different personal ratings from user B's ratings, then the system would apply less weight to his ratings for other users.