Wouldn't it be great to know who was best able to spot good ideas before anybody thought they were good ideas? It's similar to the claims of having been an early fan of a band, author, or director, or having that idea that the other person ended up building and making a lot of money/fame off of. It's closely tied with reputation--we instinctively admire people who we think can predict the future by knowing which parties to go to, which websites to build, which people to befriend, which jobs to take, which stocks to buy, etc.
I'm going to try to explain a picture I have in my head of a way to make some of this bet-generated reputation visible on the web. I'd rather talk about it figuratively rather than literally, for now, so here's a metaphor: imagine that you're watching an plant growing in time-lapse film, and it starts out a nice baby plant and slowly grows over the period of a couple years. Now, imagine a second plant next to the first orange. It's a magic plant. The second plant is will attempt to mirror (predict) how the first plant will look several months in the future. So, the way that the second plant looks is its best-guess attempt to look like the first plant in the future. Okay, and just for fun, there's a little electronic display next to the second plant which, after a couple months goes by, will tell you how well the second plant was able to predict the appearance of the first plant several months before.
The second plant has a lot of feedback to inform it in the present. Some of it is generated by the first plant: how does the first plant seem to be doing... are there any outside influences (wind, rain, animals) around that might have a long-term impact on its growth? How much stability does the environment have in general? What patterns can you find--does the plant act differently seasonally, etc? Finally, it has the electronic display to inform it on how accurate its own previous guesses were. It can learn to trust its instincts or generate new instincts if its natural instincts don't seem to be working.
The electronic display is the second plant's reputation. Now, imagine that there are a hundred magic plants like the second plant. Each one trying to predict how the first plant will be in a couple months, and each one receiving reputation based on its ability to create accurate predictions.
What will emerge from this ecosystem? Will certain magic plants emerge as better predictors? What can we learn from the good predictors to help improve the predictive powers of the other magic plants?
I can talk about it a little more literally, I guess. Because I'm getting confused. Think about a bet that has two choices:
# Seattle will be sunny tomorrow.
# Seattle will be rainy tomorrow.
Someone places their money on option 1. Another places their money on option 2. You can think of this bet as now having odds of 1:1. The next person puts their money on 1, and the odds become 2:1 that tomorrow will be sunny. Say that the next 10 people all put their money on option 1. The odds become 12:1 that tomorrow will be sunny. Finally, someone puts their money on option 2, and the final odds say that there's a 12:2 chance, or 6:1, chance of it being sunny tomorrow.
Tomorrow comes around and it is sunny. Who should gain the most reputation for their choice? I'm willing to make a case that the first person that placed their money on option 1 should gain the most reputation. When they place their bet, the odds were 0:0, or, unknown. If we treat 0:0 odds the same as 1:1 odds or 2:2 odds, they will get the same amount of money back as they put in. The last person that bet on a sunny day tomorrow, on the other hand, bet when the odds were 11:1, and therefore should only get 1/11th of the money they put in back. Both people that bet on the rainy day lose all of their money.
Say that the opposite thing happened and that tomorrow comes and it is rainy. All 12 people that bet on sunny lose 100% of their money. The first person to bet on rainy gets the same amount of money back as they put in because the odds were relatively flat when they bet. The last person to bet on rainy, however, was going up against 12:1 odds and will receive 12 times as much money back for accurately betting on rainy when it seemed least likely that it would actually be rainy.
There's some weird math going on with dividing by zero that I haven't figured out yet, but the point is that people should get rewarded based on the perception of the group at the time of prediction rather than the odds at the time that the bet has resolved. You can then create an interesting moving graph that represents how the odds of a bet change over time as additional information is revealed. In the case with the weather prediction, we're continuously getting more information about the weather, and as we get closer in time to tomorrow, it becomes easier to predict. The graph of how the odds change over time is a representation of how idea evolved over time as new information came in. Going back to the plant example, the variation in odds over time represents the information environment of the idea, like the soil and the rain and the sun that impacted the plant over time.
Why don't we create a website that lets you place bets on ideas with your reputation? Depending on when you place your bet, and for which side, you can gain or lose reputation. Then, by seeing who is able to maintain a high reputation over time you can see who's the best idea spotters out there, and begin to trust them more... if the bets are public then you may see that when Person X (who has a high reputation) bets a certain way that they tend to have an impact on the odds that bets carry. Seems like a very interesting application to me.
Recent Comments