Are trust ratings just raw sums, or can they be based on how much people you trust have rated someone?
For instance, I don't care about how many sockpuppet accounts have endorsed someone, I only care about how they're rated by people I know or trust (or, to a lesser extent, how they're rated by "friends-of-a-friend", though there should be a more than linear decay between each degree of separation.
Also, having publicly (or even privately!) viewable trust ratings can skew the ratings. I might have a true opinion of someone that's better or worse than what I'm willing to publicly announce (we can't all be Zed Shaw ;-). Viewable ratings also create the problem that I might feel pressured to rate someone well so that they won't retaliate against me: let's call this the ebay effect ;-).
There also remains the problem of proving that a given account actually belongs to the person named.
You need to have a posted, credible policy that you won't tamper with or sell ratings, or reveal private information (such as how I really think of so-and-so).
A system like this could be great, but you have to take into account every possible way it could be gamed or cracked, and build in solid defenses against all of them from the very start.
You got it right, the idea is to make these trust ratings based on how much people you trust have rated someone. In the alpha version it does not go that far yet, that's the idea though.
I agree that there has to be a good way to prevent how the system can be gamed. We are building those into our design, just haven't had the time to totally add it into the front yet.
Comments
Are trust ratings just raw sums, or can they be based on how much people you trust have rated someone?
For instance, I don't care about how many sockpuppet accounts have endorsed someone, I only care about how they're rated by people I know or trust (or, to a lesser extent, how they're rated by "friends-of-a-friend", though there should be a more than linear decay between each degree of separation.
Also, having publicly (or even privately!) viewable trust ratings can skew the ratings. I might have a true opinion of someone that's better or worse than what I'm willing to publicly announce (we can't all be Zed Shaw ;-). Viewable ratings also create the problem that I might feel pressured to rate someone well so that they won't retaliate against me: let's call this the ebay effect ;-).
There also remains the problem of proving that a given account actually belongs to the person named.
You need to have a posted, credible policy that you won't tamper with or sell ratings, or reveal private information (such as how I really think of so-and-so).
A system like this could be great, but you have to take into account every possible way it could be gamed or cracked, and build in solid defenses against all of them from the very start.
You got it right, the idea is to make these trust ratings based on how much people you trust have rated someone. In the alpha version it does not go that far yet, that's the idea though.
I agree that there has to be a good way to prevent how the system can be gamed. We are building those into our design, just haven't had the time to totally add it into the front yet.
Good suggestions. Thanks.