Having read the EXCELLENT book The Wisdom of Crowds a while back, I'm a huge proponent of doing a lot more work in this intriguing area.  Nitin Karandikar over at Read/Write Web has written a post summing up the interesting talk recently at the yahoo.confab event about using prediction markets within organizations.  This is the first I've really hard about this idea being used as widely as the enterprise, so it's exciting to see the topic talked and posted about.

The idea behind prediction markets is that if you harness the collective intelligence of a large group of a people, you'll arrive at decisions that are far more accurate than even the smartest individual in the group.  Value also plays an important role, as in the participants in the market need to have some type of vested interest in being accurate and profitable in the market--the part that makes it an actual market and not just an opinion poll.

The enterprise is a perfect place for prediction markets--this is true business intelligence.  It's far easier to incentivize participation in a company than it would be to run a prediction market for the general public.  Logistics being the main reason in my eyes--it's just a lot simpler to set up a controlled system and distribute incentives for succeeding in the market within the structure of a company than it is to create an entirely new infrastructure.  I think prediction markets within companies have the possibility of being more accurate than any opinion polls or even prediction markets that include everyone out there.  Certainly at the very least they'll be able to get up and running faster than public markets.

On an interesting side-note to the discussion, I noticed is that IBM has developed software called BRAIN that is supposed to take bias, manipulation, and hierarchy out of the predictions of small groups.  I'm not so sure this is a good idea.  I've read Mr. Surowiecki's book, and I wonder how he feels about this.  The whole point is to take 1) the unfiltered, raw predictions of a  2) a large group of people and average them to arrive at the correct answer.  The IBM solution feels like polluting the data to me; I'd be interesting in seeing the results of the software stacked up against a "straight" prediction market.

The post also mentions some public prediction markets that I wasn't aware of before:

• newsfutures - A prediction market about the news.  According to the tickers, there's a 7% chance that there will be fewer than 100,000 troops in Iraq by April 2007, a 10% chance Castro will return to power, and a 70% probability that a man will win this season of Survivor (you'll make millions with THAT gem!)
• The Yahoo! Buzz Game - A prediction market that tries to predict which technologies will do well.  Interestingly, Google Office, an as of right now-nonexistent product, is trading at twice the price of Microsoft Office, and the three game consoles (Nintendo, Sony, and Microsoft) are in a virtual dead heat.
• The World Economic Forum - Prediction markets about important issues such as Iraq, severe weather events, diseases, and crude oil.  Unfortunately (VERY unfortunately), this site doesn't seem to generate a lot of traffic.

And then there's one of the oldest and best prediction markets, which has been around for a while, the Iowa Electronic Markets, where the Democrats currently have a slight (51 to 49%) edge over the Republicans to win the 2008 presidential election.  If you've never looked at prediction markets before, you should--they're pretty fascinating.

And when you've figured out a cool new use for them, you can go check out an open source prediction market software package like zocalo.  And maybe write your congressman or congresswoman and tell them to figure out a way to make the Policy Analysis Market work so we can stop being strip searched at the airport with X-Ray back-scattering, even if it means privatizing the market.