Lorrie Faith Cranor
AT&T Labs-Research
lorrie@research.att.com
Paul Resnick
The University of Michigan
School of Information
presnick@umich.edu
ABSTRACT
An automated negotiation protocol defines the rules of a bargaining
game between automated agents – for example a negotiation between a Web
site and a Web browser. The most general protocol would permit an
unlimited number of rounds of offers and counter-offers, with offers being
either commitments or cheap talk. A more restricted protocol, with sellers
making take-it-or-leave-it initial offers, would be easier to implement
and might focus players on simpler bargaining strategies, easing analysis
of opponents’ strategies and perhaps reducing the occurrence of irrational
strategy choices due to miscalculations. We consider possible downsides
to the adoption of a more restricted protocol: might buyers or sellers,
or both, get worse outcomes with a more restricted protocol than with a
more general bargaining protocol?
In many Internet commerce applications buyers can easily achieve anonymity, limiting what a seller can learn about any buyer individually. However, because sellers need to keep a fixed web address, buyers can probe them repeatedly or pool their information about sellers with the information obtained by other buyers; hence, sellers’ strategies become public knowledge. Under assumptions of buyer anonymity, publicly known seller strategies, and no negotiation transaction costs for buyers, we find that a restricted protocol will yield the same equilibrium outcomes as a more complicated bargaining protocol. As we relax those assumptions, however, we find that sellers, and in some cases buyers as well, may benefit from a more general bargaining protocol.
The paper is motivated by the problem of designing a protocol for a Web browser to negotiate with a Web server about what information will be revealed to the Web site and how that information will be used. Such a protocol will be part of the Platform for Privacy Preferences Project (P3P), currently being developed by the World Wide Web Consortium.
An earlier version of this paper was presented at TPRC 97, in Washington DC.
Final draft of 1/99.