Distributed Commerce Transactions

Abstract

In situations where self-interested agents are interacting in an environment of distrust, commercial exchanges may be blocked due to a lack of trust. We propose a fully distributed algorithm that each agent may run to provide guarantees about the outcomes of such exchanges. The algorithm is shown in operation on two examples, one feasible and one not. Introduction In a multi-agent system with many information sources covering a vast range of topics, an agent with a query or information request might enlist the help of a number of brokers and sources in answering the query. These other agents may add value by combining results, or having better knowledge about or access to relevant sources. Information brokers may obtain documents from other sources and re-sell them to customers, who may in turn be brokering the documents to someone else. In exchange for provided services, these agents expect payment. The environment is one of distrust, so that a customer will not give payment before being certain of getting the document. Similarly, a source will not provide a document before being certain of getting paid. Trusted intermediaries (such as a shopping mall, Internet service provider, electronic bank) alleviate the problem, providing a secure way to handle a transaction. The trusted intermediary receives the document from the source and money from the customer, then performs the exchange. If one party does not provide its promised piece, the exchange is canceled, and the goods are returned to their original owner. If all of the participants could find a single trusted intermediary, the interaction would be straightforward. But these are not realistic assumptions. Moreover, the parties may have no previous history of interaction, may not exist in the same jurisdiction or even know each other's identities. Therefore, a single trusted intermediary is unlikely. More reasonable, we believe, is a series of pair-wise interactions, which taken in combination, result in the quer