Unformatted text preview: common-values 12 . We model an auction with agents that have private independent values for a good, but uncertainty about the value. We believe that this model is especially relevant in on-line auctions, where agents can have hard local problems e.g. a manufacturing agent that bids for components, or goods are non-standard and di cult to value e.g. collectibles at www.ebay.com. We assume that agents have an option to re ne their value for a cost. The cost is considered to represent the actual cost of consulting an expensive expert, or the cost that results from suboptimal or missed bids because of lost deliberation about the value of goods in other marketplaces. Conceptually, one can partition the decision problem of a bidding agent into three sub-problems: metadeliberation, valuation, and bidding. The optimal bidding strategy depends on the auction and an agent's possibly approximate solution to its valuation problem. The optimal metadeliberation strategy follows from an analysis of an agent's valuation problem and bidding strategy. An agent bids only when it has decided to perform no more deliberation about the value of the good.1 Previous models of agent-mediated markets have addressed the complexity of the bidding problem, that is deciding on an optimal bid given the value of a good, but largely ignored the valuation and metadeliberation problems although see 24, 25 .
1 In an alternative model agents do not explicitly solve the valuation problem, but select a bidding strategy directly, based on the payo from past bids 2 . It is often useful to separate the valuation and bidding phases because: a there might be a separation of skills information for example when a software agent bids for a person that values the good; b in a business context, separation can enable a bidding agent to leverage existing decision analysis tools and models for the valuation problem; c in some markets for example incentive compatible markets the bidding problem is trivial. 3 2.1 A Simple Theoretical Model
We propose a simple model for the valuation problem of an agent, that enables the derivation of optimal metadeliberation and bidding stra...
View Full Document
- Fall '13