[atlas-l] on "feedback"

[jer]

I've been thinking quite a bit about how a standardized feedback  
system might work, for Brokers to offer some kind of quality measure  
back to both the Collectors and Factories from which the results came.

One of the conclusions that I'm starting to come to is that positive  
feedback is bad.  Both in that it's prime for abuse, and in that it's  
a potential privacy issue.  The abuse factor means that it's yet  
another item that a system would have to track quality and reputation  
on, and once it's decided that the positive feedback is genuine it  
then becomes many more factors to try and deal with as each Broker  
would likely behave differently.

As for privacy, I keenly want to avoid with great distance *any* kind  
of ability for Collectors/Brokers to learn more about the user, that  
is absolutely the Brokers job and any user information should never  
leave that system.

What I do think is worth standardizing is negative feedback.  One of  
the most powerful tools an end user can have is the ability to say  
"this is wrong" and decide to share that with everyone.  The value of  
a recognizing something incorrect is immediately important to  
everyone, the Broker, Collector, and Factory.

Again, being keen on privacy, this "problem report" should be as  
generic as possible, and contain really only two items, the knugget  
ID, and a human string.  The negative feedback wont even have the  
context of the query unless it was given in the string.  Primarily,  
these problem reports should be (aggregated and then) processed by  
people within the Collectors and Factories.  A Broker may (often) use  
generic templates for the string parts, making it easier to aggregate  
them for review.

A reported knugget could be put on an automatic temporary hold list  
if there is some metric met and the reporting Broker is in good  
standing, so that this process has immediate value instead of waiting  
for admin attention.

Just some random thoughts :)

Jer