Manipulating Information Providers Access to Information in Auctions - Technologies and Applications of Artificial Intelligence - page 16

distinguish between values. This becomes possible when the information
provider's ability to provide accurate information depends on inputs received
from the auctioneer. For example, in the sale of the firm example, the board
can decide to provide the analyst with accurate, yet aggregative, informa-
tion, such that the information provider can estimate future sales as weak
or strong rather than a certain figure from a wider range of values. The non-
intuitiveness of doing this is attributed to the fact that at the end of the day
the information provider's information is offered for sale to the auctioneer
herself, thus by restricting the information provider's ability to distinguish
between values the auctioneer restricts himself by not having the choice of
purchasing more accurate information.

The paper is structured as follows. In the following section we provide a for-
mal presentation of the model. Then, we present an equilibrium analysis and
illustrate the potential profit for the auctioneer from influencing the accuracy of
the information that can be provided by the information provider. Finally we
conclude with a review of related work and a discussion on the main findings.

2 The Model

Our model considers an auctioneer offering a single item for sale to n bidders
using a second-price sealed-bid auction (with random winner selection in case
of a tie). The auctioned item is assumed to be characterized by some value X
(the “common value”), which is a priori unknown to both the auctioneer and
the bidders [16,17]. The only information publicly available with regard to X
is the set of possible values it can obtain, denoted X ∗ = { x 1 , ..., x k } ,andthe

probability associated with each value, Pr ( X = x )( x∈X ∗ Pr ( X = x )=1).
Bidders are assumed to be heterogeneous in the sense that each is associated
with a type T that defines her valuation of the auctioned item (i.e., her “private
value”) for any possible value that X may obtain. We use the function V t ( x )
to denote the private value of a bidder of type T = t ifthetruevalueofthe
item is X = x . It is assumed that the probability distribution of types, denoted
Pr ( T = t ), is publicly known, however a bidder's specific type is known only to
herself.
The model assumes the auctioneer can obtain information related to the value
of X from an outside source, denoted “information provider”, by paying a fee
C that is set by the information provider. Similar to prior models (e.g., [34]),
and for the same justifications given there, it is assumed that the option of pur-
chasing the information is available only to the auctioneer, though the bidders
are aware of the auctioneer's option to purchase such information. In its most
general form, the information provided by the information provider is a subset
X ↂ

X ∗ , ensuring that one of the values in X isthetruecommonvalue.This
is usually the case when the information provider cannot distinguish between
some of the possible outcomes however can eliminate others. Therefore, the in-
formation provider will provide a subset X ∈ D = { X 1 , ..., X l } where D is the