A learning-enabled infrastructure for electronic contracting agents

Abstract

With the vigorous development of electronic commerce these years, many experts and scholars have devoted themselves to various fields of research and application. Of these fields, electronic contracting is a new research topic in great demand. In spite of its promise, electronic contracting involves the standardization of ontology and automation of negotiation, which renders the implementation of electronic contracting difficult. In view of the necessity of electronic contracting, we present a learning-enabled agent-based infrastructure and claim that it will be a solution to the problems encountered during the process of electronic contracting by a variety of evaluations. In this infrastructure, the applications of an agent are viewed as a set of application ontologies, each of which is a combination of a context ontology and a object ontology so that the negotiation context and automation of negotiation can be flexibly integrated in this infrastructure. The infrastructure enables the automation of electronic contracting through a general and automatic communication protocol and provides reusability by the componentization of agents. The infrastructure provides personalized multiattribute evaluation and proposal generation by a mechanism, which is a combination of neural networks and genetic algorithms, in order to enable the automatic negotiation ability at agents.

Introduction

With the advent of vigorous development of electronic commerce in recent years, many researchers and practitioners have devoted themselves to various fields of research and application. Of these fields, electronic contracting is one of the most important research topics, and is in great demand.

Baum and Perrit (1991) gave a definition of electronic contracting: ‘Electronic contracting involves the exchange of messages between buyers and sellers, structured according to a prearranged format so that the contents are machine-processible and automatically give rise to contractual obligations.’ On the other hand, Runge (1998) claimed that electronic contracting involves two processes, agreement negotiation and agreement signing. Agreement negotiation is the exchange of information between buyers and sellers according to their negotiation, and agreement signing is the signing of contract on the basis of the negotiation results.

Accordingly, there are three important processes for electronic contracting:

• The communication is based on a common protocol.

• The means to negotiation and cooperation is (semi-)automatic.

• The signing and execution of the contract is based on the negotiation results obtained electronically.

The rationale behind the great demand of electronic contracting can be realized by the following observations:

• Runge (1998) brought up a survey report about electronic contracting: ‘The majority of users in the WWW regard electronic contracting as important and would be willing to negotiate all kinds of contract conditions within the first 15 min by themselves.’

• According to the consumer buying behavior model (Guttman, Moukas & Pattie, 1998a), consumers must perform four tasks: need identification, production brokering, merchant brokering, and negotiation in the space of time. A useful system that can support electronic contracting is indispensable to consumers in order to accomplish the four tasks in time.

• Dutta and Segev (1999) discussed how businesses can be transformed in the marketplace from the angles of four P's (Product, Price, Promotion, and Placement) and one C (Customer relationship). These transformations involve dynamic customization of the 4P's, online negotiation, etc. So far, only 12% can be the achieved on the rate of the transformation of price for some enterprises, and the rate in most enterprises have been lower. A flexible and efficient electronic contracting infrastructure, we believe, is the solution to the problems of the transformation of businesses.

In spite of its promise, electronic contracting involves the standardization of ontology and automation of negotiation, which render the implementation of electronic contracting difficult. Standardization of ontology makes it possible that buyers and sellers can exchange information of their interests without the problem of negotiation context understanding. Automation of negotiation is the virtue of electronic contracting, in which contents are machine-processible and automatically give rise to contractual obligations. In recent years, CommerceNet Consortium (Glushk, Tenenbaum & Meltzer, 1999) have devoted themselves to the task of the standardization of ontology. Automated negotiation also has been investigated with different perspectives as described in Section 2. However, there are still drawbacks with existing approaches (for details, please see Section 2):

• Electronic contracting agents can not be reused easily.

• Automated negotiation either is price-dominated or presumes linearity among attributes of targets that are being negotiated.

• The negotiation context and automation of negotiation are not flexibly integrated for the provision of electronic contracting capabilities.

In view of the shortcomings of the existing electronic contracting approaches, it naturally leads to the question—what approach should be employed for electronic contracting agents in order that reusability, non-linearity multiattribute customized negotiation, and flexible integration of context capability and negotiation capability can be reached at agents? In this paper, we present a learning-enabled agent-based infrastructure, Agent Electronic Contracting Wrapper (Agent EC-Wrapper), which serves as an answer to the problems encountered during the process of electronic contracting.

Since the signing and performing of a contract are essential to electronic commerce (Runge, 1998), and there have been many remarkable research studies about it, we will focus this paper on these first two processes of electronic contracting. Furthermore, considering that (multi)agent-based electronic commerce is in vogue, we believe that a general agent-based infrastructure, including conversation protocols, ontology of products and services, and support of negotiation, will enable multiagent system to make electronic contracting on a common infrastructure.

The rest of the paper is organized into six sections. In Section 2, we discuss some relevant works. Section 3 presents the infrastructure, an Agent EC-Wrapper, for electronic contracting agents, and Section 4 describes the components of this infrastructure. In Section 5, we give an example to demonstrate the capabilities of this implemented infrastructure and discuss three directions of applications on which the infrastructure can be applied. Section 6 then evaluates this infrastructure in terms of efficiency and quality. Finally, a conclusion is made in Section 7.