Brokering in electronic insurance markets

Abstract

Almost all insurers now have a web site providing information on the company, its products and contact details, but the scope of insurance activities potentially affected by e-commerce is much broader than that, giving rise to a host of interesting issues. Also, customers buying over the Internet look for added value. This paper proposes a distributed multi-agent system for e-insurance brokering where customers are grouped together, exploiting user modelling and machine learning techniques, as an approach to better match customers and insurance product offers. Particular attention is paid to the interpretation of the generated communities. For this purpose, we use a metric to identify the representative insurance product configuration of each community. To improve broker's evaluation of received insurers' bids we propose the automatic construction of the customer's profile, reflecting its preferences on all the attributes of an insurance product.