SmartVote: a really interesting idea

This blog post  http://garthzietsman.blogspot.com/2011/10/smart-vote-concept.html has a really cool idea in it. I saw this post referred from  http://lesswrong.com/r/discussion/lw/cwm/link_freakostats_and_cev/

To describe the idea, let me first mention another idea, which is prediction markets.  Here, you allow people to "buy" an interest in the outcome to some question, like "Romney will win the US presidential election in 2012" or "At the end of 2013, oil will be selling for more than $150/barrel."  Essentially you can buy a share that will yield some fixed amount, perhaps $10 if the answer is "yes."  You can also "sell" that same share (or conceptually, buy a share that the answer is no).  The market clears: if more people want to buy shares of yes than sell them, then perhaps the price rises to $8 a share before the buyers think it is too expensive to keep buying.  Then at any point, an external observer looking at activity in the market can declare that the market for an $8 price to get a $10 return on yes is saying "there is an 80% chance the answer is yes."  

The important idea with a prediction market is that the market is more correct on average than are any of its participants.  Of course its participants are making money only to the extent that they are more correct than are other participants in the market (as I have described it, it is a zero-sum market).  But the utility of the market is supposed to be for those who look at the "collective wisdom" there and get better estimates of how things will go than they could get if they just hired one or two of the smart participants in the market.  

Enter Garth Zietsman and his "SmartVote" concept.  Dr. Zietsman talks about research results showing that in a bunch of tests, he can get the correct answers by comparing "smart" (high IQ) voters against less smart voters.  In at least some of these tests, where even the smart people averaged 42% correct on these tests, the simple analysis of the multiple smart answers vs the multiple less smart answers gives 100% correct answers.  

Intuitively, this makes sense to me.  Products are much more complex than any one mind can comprehend, and they are all created by distributing the intelligence work across many people, and having many layers of abstraction and simplification in order to coordinate across parts off the product.  If human intelligence in design can be multiplied by ganging together many minds, than why couldn't many minds ganged together do better on a test?  

Very intriguing stuff in my opinion.