The Text-Authoring Market
We have already seen that a prediction market can be used to approve a text. Predictive cost-benefit analysis, after all, predicts whether a regulation will have more benefits than costs, a condition that might count as constituting approval. This approach can be easily generalized. Given a requirement to approve a text for a particular purpose, anyone might be allowed to propose a text. A prediction market could then be used to anticipate the evaluation of an ex post decision maker who will decide whether the text should have been approved or disapproved. If the prediction market anticipates that the text should be approved, then it is approved. Otherwise, other participants can propose alternative texts, perhaps with the requirement that the individuals pay the subsidy for the market. Eventually, this market might well approve a text. A publisher might use it, for example, to decide whether to publish novels submitted to it. Although one or more humans will have written each proposed novel, the market acts to approve texts, and so we might count it as a text-authoring market. This approach, however, leaves much to be desired. Ideally, there should be an opportunity for multiple human authors to coordinate their activities and produce a jointly written text. What is needed, in short, is a means for individuals to make edits to the texts of others. This modification is relatively easily accomplished. Initially, the market’s text is blank, but anyone can propose amendments to it. Whether an amendment to the text is accepted would be determined by a prediction market anticipating the evaluation (which might occur only with some positive probability) of an ex post decision maker. In effect, each amendment replaces the proposed text with another text. Amendments also might be proposed to amendments. Once the text has been fully amended, a prediction market can be used to determine whether to accept the text as a whole. This approach assures that the creation of the text is a collaborative effort of all market participants. And yet it need not be they whose taste determines what is included. Rather, it is the participants’ prediction of the taste of the ex post decision maker. And so, by altering the pool of ex post decision makers, one can alter the incentives of the authors. The principal complication concerns timing. If we allow anyone to offer an amendment to any text and provide a set time for the market to consider it, then the market might never get around to approving a text, especially if there are some market participants who would benefit from such delay. We thus need some mechanism for ending prediction market deliberation about individual amendments and about the text as a whole. Prediction markets, of course, can be used to make this assessment. At any time, someone might call for resolution of a decision on a particular amendment, and a prediction market would be used to assess this call for a decision. If that market predicted that an ex post decision maker would agree that it was time for a decision, then the amendment would be accepted or rejected in its current form, and all undecided proposed amendments to the amendment would be rejected. The same approach could be applied to the text as a whole. There are, of course, possible variations on this system. Instead of relying on repeated separate markets for assessing whether the market should end at a particular time, a single continuous market might be used. For example, the market might predict an ex post decision maker’s assessment of the optimal ending time for the market. The market would then end once its prediction of the optimal time exceeded the actual time–if that prediction continued to hold for some period of time. That period of time would ideally be partly randomized and would need to be long enough to allow participants to discipline any attempts at manipulation by those who have an interest in the market’s ending, so that no participant could unilaterally force an end to a market by effecting a transient change in price. Once the market t ended, the ex post decision maker would assess when the market optimally should have ended. Anticipating this decision, market participants would try to determine what amount of time made sense for a particular market, updating assessments based on new developments. Market participants presumably also would consider the relations among proposed amendments. For example, ordinarily they would decide that an amendment to an amendment should be resolved before the underlying amendment, though sometimes they might conclude that it makes sense to resolve an amendment without considering an amendment to it. The only important remaining issue is how to subsidize the text-authoring market. It may be important to subsidize both those who make predictions and those who suggest text and amendments. First, both the main prediction markets and the timing markets would receive some subsidy. The level of the subsidy might be determined by yet more prediction markets, which could forecast an ex post decision maker’s assessment of the importance of each of the main prediction markets and the timing prediction markets. Each text-authoring market as a whole might be given a single fixed subsidy. This fixed subsidy could be distributed to market participants and amendment proposers in proportion to the assessments of the subsidy prediction markets. Of course, the market designer would also need to subsidize each of the subsidy prediction markets. It would be possible to use prediction markets to perform that task, too, leaving the market designer the task of setting a subsidy for the markets that determine the subsidy of the main and the timing markets. Second, each text proposal could receive a subsidy, too, perhaps a portion of the single fixed subsidy, with the exact portion to be determined by a normative prediction market. The subsidy for a particular text proposal also could be determined by a subsidy prediction market to gauge the assessments of ex post decision makers. They presumably would recommend higher subsidies for successful amendments, although they might find that some unsuccessful ones deserve a subsidy as well, for example, if they inspire some other amendments.
One Response to “The Text-Authoring Market”
Leave a Reply
January 30th, 2008 at 10:02 pm
[…] agree that field tests comparing these with alternatives would be useful. Two of the proposals (the text-authoring market and the market web) could certainly benefit from experimentation, but the software needed to […]