Search

This dissertation explores a new class of intelligent software agents called product selection agents (PSA) and proposes a business model based on the concept. A product selection agent is an intelligent software agent, coupled with comprehensive databases, that performs product selection and associated functions on behalf of its clients. This dissertation investigates properties these agents must have in order to fulfill functions suggested by the literature in consumer decision-making and by trends in e-commerce. We also provided rationale for the development of product selection agents. We present a new typology of decision scenarios explicitly for the world of e-commerce. In the typology, the most difficult decision scenario is to select products that are best for the client who lacks well-developed preferences and who does not have the ability to construct preferences priori to a decision. Through a series of experiments we show that the Support Vector Machine can be applied to overcome design barriers and fulfill the goals of the client.To illustrate the design principles, we discuss how to build an agent for the pharmaceutical industry, labeled Case 4, in the most difficult context. The PSA will help diabetics select glucose meters that best meet their needs, particularly as these needs change as a function of stage of diseases, life style, demographics, and the client's economic status. The agent will also serve physicians specializing in the treatment of diabetes, most of whom cannot keep current about products on the market, their functioning, their costs, and other aspects regarding the appropriateness of a given product or product bundle for a given patient. Further, because all participants in the industry share the agent through the Internet, they will be able to exchange information in real time, thus integrating the industry using the agent as infomediary.