[Foundation-l] Rewarding volunteers

Delirium delirium at hackish.org
Sun Sep 12 18:44:56 UTC 2004


Jimmy (Jimbo) Wales wrote:

>Algorithmic reputation systems are always a temptation, just because
>they are possible.  But human reputation systems are much more nuanced
>and powerful.
>  
>
I wouldn't mind a combination of the two---for example, being able to 
say "I think so-and-so has been doing a very good job lately".  
Wikipedia is getting big enough that I can keep track of a handful of 
people I personally think are doing a good job, but even finding out 
who, say, the 10 people I trust most themselves trust is a daunting 
task, and one that could be simplified by algorithms.  Then I could have 
information like "I don't know this person myself, but it looks like a 
large number of the people I trust trust him/her", which could be useful.

Even slightly more complex systems, like having a few multiple 
ratings---perhaps one for factual knowledge, one for neutrality, one for 
quality writing, and one for wikiquette---would be useful IMO.

Note that I'd propose this _not_ as a global rating, but as a very 
user-specific rating, with hopefully a small horizon: it would give me 
ratings based on what I personally rated people, and what the immediate 
people I rated did (probably not more than two degrees of separation).  
It wouldn't give a rating like "so and so is rated excellently in 
general".  So in that sense, it wouldn't solve the "reward" problem that 
was proposed here, but would solve a different problem---I used to know 
all the regular Wikipedians (and even review almost all of each day's 
edits and new pages!), but that is nearly impossible these days, so some 
algorithmic way to help me deal with the information overload would be 
helpful.  Perhaps "show me all new pages created by people who are not 
known either by me or by the people I know," and various other such filters.

-Mark




More information about the foundation-l mailing list