A Cure?

By | March 5, 2006
2 Comments

Michael Strong opines that the solution to some of these dangerous, intoxicating delusions that we’ve been talking about — at least to the extent that they show up in academia — might be predicition markets:

Prediction markets may provide a powerful antidote to academic abuses of power. Just as critical experiments have enforced a reality-based ethos of scientific integrity in the hard sciences, so, too, prediction markets may be used to enforce a reality-based ethos of scientific integrity among those who make claims about society, politics, and economics. When theoretical speculations in the hard sciences repeatedly produce false predictions they are marginalized. We need to create just such a system for the social sciences.

For those of us who believe that excessive respect for academic opinion has led to avoidable poverty, waste, and human misery due to adherence to the misguided policies proposed by academics in the past century, then the need to create a more reliable system for discovering the empirical truth about the political and social realm is urgent. Moreover, prediction markets have the added benefit of allowing ordinary people to make money, possibly fortunes, by exposing those unfounded academic beliefs that are most widely accepted at present. Society wins, ordinary people win, and we create an incentive for people to focus their energy and attention on discovering important facts about the future.

Sounds like a win-win to me.

In addition to assessing the probabalilities of various outcomes, I think predicition markets may eventually incorporate more elaborate descriptions of the future. Compare a potential future as described by the Foresight Exchange with a different future as described by Long Bets. Foresight — along with most other currently operating predicition markets — is currently interested primarily in the ouctomes. Long Bets, on the other hand, is as interested in the arguments supporting the outcome as it is in the outcome itself. However, even the Long Bets approach is binary. The thing either will happen or it won’t, so an argument is provided for each side.

In fact, there are many different sets of circumstances and events that can lead to the same outcome. I’d like to see a prediction market that rewards both correct answers and the best analyses as to why an outcome will occur.

  • Karl Hallowell

    Given the value of predictive markets, I’d like to point out a danger with the recent attempts in the US to outlaw online gambling. This wouldn’t affect play money sites like the Foresight Exchange, but it certainly would damage real money sites like TradeSports, which in addition to a large amount of not very relevant sports betting, bets on political, financial, and other events (like the capture of Osama bin Laden) that have general relevance.

    Further, the current state of these markets are often reported in the media. I think that points out one of the unintended consequences of the online ban. Namely, that it hinders an important source of information for society.

  • Karl Hallowell

    One way I’ve heard of modeling “why” is to model scenarios. Robin Hanson and others have a large body of work on the subject.

    The key is correlations. The market needs to have a relatively low capital way to signal various correlations between the underlying finite state hypotheses. For example, the Foresight Exchange allows for some modest correlation. After all, if claim A rises, that may indicate to you (based perhaps on your knowledge or on past trading behavior you have observed) that the price of claim B should fall. Currently such trading would accomplished by selling both claim A and claim B or more risky, just by selling one of the two claims.

    An additional step would be to differentiate on when the claims are settled as well. Ie, if a model states that event A will produce event B at a later time, then a rise in A would indicate that the trader should bid up B (lagged in time by the appropriate amount).

    One problem here is that the more details you add, the less likely it is for the market to remain liquid, ie, to have enough traders in it to generate useful information and to support trading for nonspeculative purposes (like hedging).

    Dr. Hanson had an ingenious solution in terms of scoring rules for a polling-like system. The idea is that the market has a certain polling value for a large number of hypotheses. You “correct” these hypotheses by submitting a new set of values for the hypotheses. Each such action incurs a cost. If your correction is substantial, then it has a large cost associated with it.

    As I understand it, something along these lines was proposed for the “Policy Analysis Market” which was a DARPA project. Unfortunately, the project was discovered by several uninformed (or perhaps malicious) senators who characterized it as a “terror” market which would trade in contracts about whether terrorist attacks would occur. Apparently, the theory was that terrorists would trade directly in this market so as to profit from their advanced knowledge, and it would hence encourage terrorism.

    This was the event that lead to the resignation of John Poindexter and the craven knuckling under of the Department of Defense.

    This has turned into yet another mini-rant on the myopic politics surrounding prediction markets and gambling in general. Some day, the US (and collectively the world) will need to mature and face the fact that betting whether legal or illegal is widespread and embrace the power of predictive markets.