Socionics "rating agency"

[Marie84]

I think everyone pretty much has some sort of internal filter to choose who they see as reliable and not without needing a popularity system instated, which doesn't actually prove anything.
A more useful indicator could be to have users add somewhere the approach that they use, so that users who are less knowledgeable with the members here and have a preference for one school of thought over another can differentiate who to take more seriously.
Although I think eventually most people catch onto this naturally

[Bluenoir]

This sounds like a completely arbitrary popularity contest.

Define a core group in which everybody knowns everybody IRL and let them type the rest of the community, or something

What for, and why should anyone accept this core group as an authority?

To "reward" members with views that conform, and try to bring fringe opinions in line

The inquistion is upon us!

[Jonathan]

The Russians have already figured this out. They rate people based on how well their typings correlate with the group as a whole. It would also be good to subdivide it via some sort of factor analysis or cluster analysis....because one would find that there are certain "schools" where some people within that school have a high correlation.

Another very interesting way to do it is with multidimensional scaling. Based on people's ratings of a set group of people, dots representing the raters could be placed on a 2-dimensional map where people who rate similarly will appear next to each other. People who are more "mainstream" in their typings will probably appear in the center. There's software to do this.

The point is, only such statistical approaches which measure similarities between raters would work....because as soon as you appoint an "authority," that skews things because probably the authority is wrong (or many people will disagree" and then the authority will sanction only people within that "school."

The other approach is to do studies on how well people's ratings adequately predict intertype relations, but that's rather time-consuming and complicated to implement.

[krieger]

it seems to me you could pretty much deduce mechanically what one person's rating is toward another based on how many of their typings overlap. in the end it would not add any new information beyond the influence of silly drama, squabbles and favoritism.